A method, apparatus and vehicle for controlling a vehicle door

Abstract

This application discloses a method, device, and vehicle for door control. It predicts collision information when the door collides with an obstacle within its range, and generates a risk vector based on the collision information to accurately predict the positions of vehicles approaching from the side and rear, pedestrians, and surrounding obstacles. Simultaneously, it introduces confidence-based decision-making to improve robustness in complex scenarios and effectively handle uncertainties such as sudden pedestrian changes of direction or electric vehicles accelerating to overtake. Based on the handle grip force signal, occupant posture information, and seat information inside the vehicle, it generates an internal intent vector to accurately identify the user's intention to open the door. The opening resistance of the door is adjusted in real time according to the risk vector and the internal intent vector, abandoning rigid locking logic. This satisfies the user's need to open the door and observe, while also intervening promptly in high-risk situations, improving user comfort and acceptance, and achieving a dynamic balance between safety and experience.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to a method, apparatus and vehicle for door control. Background Technology

[0002] When a vehicle is parked, collisions with vehicles, obstacles, or pedestrians to the side or rear are a frequent and high-risk scenario when occupants open the car door. There are blind spots when opening the door and observing the situation to the side and rear through the rearview mirror, especially in poor lighting conditions or when the vehicle is parked close to a lane or obstacle, making it difficult to control the maximum opening angle and safe distance of the door.

[0003] In existing technologies, audible / visual alarms can be used to alert people opening car doors to a potential collision risk from the side and rear, allowing them to open the door only after the risk has passed. However, this alert is weak, and people opening the door may easily ignore it or not have enough time to react. Alternatively, the doors can be temporarily locked and unlocked only after the risk of a side-rear collision has passed. This method is safe, but the intervention is abrupt, and delayed unlocking can lead to a very poor user experience. Summary of the Invention

[0004] In view of the above problems, this application provides a method, device and vehicle for door control, so as to detect the risk of collision when opening the door in advance, and control it in a way that is free from rigid intervention and has a good user experience, so as to avoid collision.

[0005] This application discloses a method for controlling a vehicle door, the method comprising: The system predicts collision information when a door of a target vehicle collides with an obstacle; the obstacle is located within a preset range of the target vehicle and includes static and dynamic obstacles; the collision information includes the limit opening angle when the door collides with the static obstacle, the collision time between the door and the dynamic obstacle, and the prediction confidence level corresponding to the dynamic obstacle. A risk vector is generated based on the collision information; the risk vector is used to characterize the risk of the vehicle door colliding with the obstacle. An internal intent vector is generated based on the in-vehicle information of the target vehicle; the in-vehicle information includes handle grip force signal, driver and passenger posture information, and seat information; the internal intent vector is used to represent the intention of the person opening the car door. The opening resistance of the vehicle door is controlled based on the risk vector and the internal intent vector.

[0006] Optionally, the collision information when the predicted target vehicle door collides with an obstacle includes: Based on the three-dimensional contour model of the static obstacle, the maximum opening angle is predicted; Predict the trajectory of the dynamic obstacle; The collision time is calculated based on the motion trajectory; The prediction confidence level is determined based on the motion state, historical motion trajectory, and category of the dynamic obstacle.

[0007] Optionally, predicting the ultimate opening angle based on the three-dimensional contour model of the static obstacle includes: The vehicle-mounted sensing device scans the preset range when the target vehicle stops, and obtains scanning data; the vehicle-mounted sensing device includes a camera and an ultrasonic radar. A point cloud map containing the static obstacles is constructed based on the scan data; The three-dimensional contour model of the static obstacles in the point cloud map is established; Based on the configuration information of the car door, the distance between the car door and the three-dimensional contour model at each opening angle is calculated; The maximum opening angle is predicted based on the distance.

[0008] Optionally, determining the prediction confidence level based on the motion state, historical trajectory, and category of the dynamic obstacle includes: The motion state, the historical motion trajectory, and the category are input into a pre-trained confidence evaluation model to obtain the predicted confidence output by the confidence evaluation model; the motion state includes the rate of change of speed and the rate of change of direction; the category includes pedestrians, motor vehicles, and non-motor vehicles.

[0009] Optionally, generating an internal intent vector based on the in-vehicle information of the target vehicle includes: A handle intention vector is generated based on the handle grip force signal; the handle intention vector is used to characterize the intensity of the person opening the car door's intention to open the door. A visual intent vector is generated based on the posture information of the driver and passengers; the visual intent vector is used to characterize the degree of readiness of the person opening the car door to get out of the car. Feature extraction is performed on the collected seat information to obtain the seat's pressure center, pressure displacement trend, and pressure displacement velocity. Based on the pressure center, the pressure displacement trend, and the pressure displacement velocity, the disembarkation intention vector is generated; the disembarkation intention vector is used to characterize the intensity of the person opening the car door's intention to get out of the car. The handle intention vector, the visual intention vector, and the alighting intention vector are fused to generate the internal intention vector.

[0010] Optionally, generating a handlebar intention vector based on the handlebar grip force signal includes: Feature extraction is performed on the collected handlebar grip force signal to obtain the average grip force and grip force rise rate; If the average grip force is less than a preset grip force range and the grip force rise rate is less than a preset rate range, the handle intention vector is determined to be weak. If the average grip force is within the preset grip force range and the grip force increase rate is within the preset rate range, the handle intention vector is determined to be normal. If the average grip force is greater than the preset grip force range and the grip force increase rate is greater than the preset rate range, the handle intention vector is determined to be strong.

[0011] Optionally, generating a visual intent vector based on the occupant's posture information includes: Feature extraction is performed on the collected posture information of the driver and passengers to obtain the head turning angle, torso twisting angle, hand position, and gaze focus area; Based on the head turning angle and the focal area of ​​vision, determine the degree of observation of the target vehicle's rearview mirror / side window by the driver / passenger. Based on the torso twist angle and the hand position, determine the tendency of the driver / passenger to make an exit action; The visual intent vector is generated based on the degree of observation and the trend.

[0012] Optionally, controlling the opening resistance of the vehicle door based on the risk vector and the internal intent vector includes: A risk coefficient is determined based on the risk vector; the risk coefficient is used to characterize the degree to which the opening resistance is controlled. The basic damping component is determined based on the current opening angle of the door; the basic damping component is used to simulate the feel of opening the door. The interaction damping component is determined based on the risk vector and the internal intent vector. The opening resistance is obtained by weighting and fusing the basic damping component and the interactive damping component using the risk coefficient.

[0013] Based on the above-mentioned door control method, this application also discloses a door control device, including: a prediction unit, a risk generation unit, an intent generation unit, and a door control unit; The prediction unit is used to predict collision information when the door of the target vehicle collides with an obstacle; the obstacle is located within a preset range of the target vehicle and includes static obstacles and dynamic obstacles; the collision information includes the limit opening angle when the door collides with the static obstacle, the collision time between the door and the dynamic obstacle, and the prediction confidence level corresponding to the dynamic obstacle; The risk generation unit is used to generate a risk vector based on the collision information; the risk vector is used to characterize the risk of the door colliding with the obstacle. The intent generation unit is used to generate an internal intent vector based on the in-vehicle information of the target vehicle; the in-vehicle information includes handle grip force signal, driver and passenger posture information, and seat information; the internal intent vector is used to represent the intention of the person opening the door to open the door. The door control unit is used to control the opening resistance of the door based on the risk vector and the internal intent vector.

[0014] Optionally, the prediction unit includes: An angle prediction subunit is used to predict the limit opening angle based on the three-dimensional contour model of the static obstacle. A trajectory prediction subunit is used to predict the motion trajectory of the dynamic obstacle; A time prediction subunit is used to calculate the collision time based on the motion trajectory; The confidence prediction subunit is used to determine the prediction confidence based on the motion state, historical motion trajectory and category of the dynamic obstacle.

[0015] Optionally, the angle prediction subunit includes: A scanning subunit is used to scan within a preset range when the target vehicle is stopped using an onboard sensing device to obtain scanning data; the onboard sensing device includes a camera and an ultrasonic radar. A point cloud construction subunit is used to construct a point cloud map containing the static obstacles based on the scanned data; The model building subunit is used to build the three-dimensional contour model of the static obstacles in the point cloud map; The distance calculation subunit is used to calculate the distance between the car door and the three-dimensional contour model at each opening angle, based on the configuration information of the car door. The limit prediction subunit is used to predict the limit opening angle based on the distance.

[0016] Optionally, the confidence prediction subunit includes: The confidence determination subunit is used to input the motion state, the historical motion trajectory, and the category into a pre-trained confidence evaluation model to obtain the predicted confidence output by the confidence evaluation model; the motion state includes the rate of change of speed and the rate of change of direction; the category includes pedestrians, motor vehicles, and non-motor vehicles.

[0017] Optionally, the intent generation unit includes: A handle intention generation subunit is used to generate a handle intention vector based on the handle grip force signal; the handle intention vector is used to characterize the intensity of the person opening the car door's intention to open the door. A visual intent generation subunit is used to generate a visual intent vector based on the posture information of the driver and passengers; the visual intent vector is used to characterize the degree of preparation of the person opening the car door for getting out of the car. The feature extraction subunit is used to extract features from the collected seat information to obtain the seat's pressure center, pressure displacement trend, and pressure displacement velocity. The vehicle exit intention generation subunit is used to generate the vehicle exit intention vector based on the pressure center, the pressure displacement trend, and the pressure displacement velocity; the vehicle exit intention vector is used to characterize the intensity of the person opening the car door's intention to get out of the vehicle; The fusion subunit is used to fuse the handle intent vector, the visual intent vector, and the alighting intent vector to generate the internal intent vector.

[0018] Optionally, the handle is intended to generate a sub-unit, including: The grip force extraction subunit is used to extract features from the collected handlebar grip force signal to obtain the average grip force and grip force rise rate. The weak determination subunit is used to determine that the handle intention vector is weak when the average grip force is less than a preset grip force range and the grip force rise rate is less than a preset rate range. The normal determination subunit is used to determine that the handle intention vector is normal when the average grip force is within the preset grip force range and the grip force rise rate is within the preset rate range. A strongly determined subunit is used to determine the handle intent vector as strong when the average grip force is greater than the preset grip force range and the grip force rise rate is greater than the preset rate range.

[0019] Optionally, the visual intent generation subunit includes: The posture extraction subunit is used to extract features from the collected posture information of the driver and passengers, and obtain the head turning angle, torso twisting angle, hand position and gaze focus area. The observation and determination subunit is used to determine the degree of observation of the driver / passenger of the target vehicle's rearview mirror / side window based on the head turning angle and the focal area of ​​the line of sight. The trend determination subunit is used to determine the trend of the driver / passenger making an exit action based on the torso twist angle and the hand position; A vector generation subunit is used to generate the visual intent vector based on the observation level and the trend.

[0020] Optionally, the door control unit includes: A coefficient determination subunit is used to determine a risk coefficient based on the risk vector; the risk coefficient is used to characterize the degree to which the opening resistance is controlled. A basic damping determination subunit is used to determine the basic damping component based on the current opening angle of the door; the basic damping component is used to simulate the feel of opening the door. An interactive damping determination subunit is used to determine the interactive damping component based on the risk vector and the internal intent vector. The weighted fusion subunit is used to weight and fuse the basic damping component and the interactive damping component using the risk coefficient to obtain the opening resistance.

[0021] Based on the above-mentioned door control method, this application also discloses a vehicle, including: a processor, a memory, and a system bus; The processor and the memory are connected via the system bus; The memory is used to store a program, the program including instructions that, when executed by the processor, cause the processor to perform the steps of any of the methods described above.

[0022] This application discloses a method, device, and vehicle for door control. It predicts collision information when a target vehicle door collides with an obstacle within a preset range, including the maximum opening angle when colliding with a static obstacle, the collision time with a dynamic obstacle, and the prediction confidence level corresponding to the dynamic obstacle. A risk vector is generated based on the collision information to accurately predict the positions of vehicles approaching from the side and rear, pedestrians, and surrounding obstacles. Confidence-based decision-making is introduced to improve robustness in complex scenarios and effectively handle uncertainties such as sudden pedestrian changes of direction and electric vehicles accelerating to overtake. An internal intent vector is generated based on the handle grip force signal, occupant posture information, and seat information to accurately identify the intention of the person opening the door. The door opening resistance is adjusted in real time according to the risk vector and internal intent vector, abandoning rigid locking logic. This satisfies the user's need to observe when opening the door while also intervening promptly in high-risk situations, improving user comfort and acceptance, and achieving a dynamic balance between safety and experience. Attached Figure Description

[0023] 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 only embodiments of this application. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0024] Figure 1 This is a schematic flowchart of a door control method disclosed in an embodiment of this application; Figure 2 This is a schematic flowchart of another door control method disclosed in an embodiment of this application; Figure 3 This is a schematic diagram of the structure of a door control device disclosed in an embodiment of this application. Detailed Implementation

[0025] 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, and 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.

[0026] Example 1: This application discloses a method for controlling vehicle doors.

[0027] For details, please refer to Figure 1 The door control method disclosed in this embodiment includes the following steps: Step 101: Predict the collision information when the target vehicle's door collides with an obstacle.

[0028] In this embodiment, the target vehicle can utilize onboard sensing devices such as surround-view cameras, ultrasonic radar, and corner radar to scan the environment within a preset range around itself before or at the moment of stopping, acquiring static and dynamic obstacles near the target vehicle. The surround-view cameras provide visual texture information, while the ultrasonic radar and corner radar provide distance information. The method in this embodiment does not specifically limit the number and type of onboard sensing devices, as long as they can accurately and comprehensively perceive the environment around the target vehicle. As an example, four surround-view cameras, twelve ultrasonic radars, and several corner radars can be used.

[0029] In the method of this embodiment, the collision information may include the limit opening angle when the door collides with a static obstacle, the collision time between the door and a dynamic obstacle, and the prediction confidence level corresponding to the dynamic obstacle. Specifically, to obtain the limit opening angle, a Simultaneous Localization and Mapping (SLAM) algorithm is first used to fuse the scanning data from the camera and radar to construct a point cloud map containing static obstacles (such as pillars, walls, and neighboring vehicles) near the target vehicle. Then, various static obstacles in this point cloud map are identified and segmented to obtain the category of each static obstacle (such as "cylindrical pillar," "curbstone," etc.), and a three-dimensional contour model of each static obstacle is established.

[0030] In the method of this embodiment, after obtaining the three-dimensional contour model, the distance between the door and the three-dimensional contour model at each opening angle can be calculated by combining the configuration information such as the geometric dimensions of the door and the position of the rotating hinge, and the limit opening angle can be predicted based on the distance. For example, a function is pre-calculated and stored for the current door configuration information. This function describes the distance between the edge of the door and the surfaces of all identified three-dimensional contour models at any opening angle of the current door. For example, the rotation radius of the door is 80cm. When its opening angle is 25°, it is 5cm away from the side of a cylindrical pillar with a diameter of 30cm. When its opening angle is 26°, it is -1cm away from the side of the cylindrical pillar (or there is a 1cm scrape with the side of the cylindrical pillar). That is, when the opening angle does not exceed 25°, there is no risk of collision with the door. When the opening angle exceeds 25°, there is a risk of collision with the door, and the depth of the collision (scratching) increases non-linearly with the increase of the opening angle.

[0031] In this embodiment, vehicle-mounted perception devices such as millimeter-wave radar and side-view cameras are used to continuously detect the types of dynamic obstacles (pedestrians, motor vehicles, and non-motor vehicles) behind and to the side of the target vehicle. Target tracking algorithms, such as Kalman filtering, can be employed to assign a unique identifier to each dynamic obstacle and continuously record its trajectory. Subsequently, a deep learning prediction network (such as Social-LSTM or Transformer) is used to predict the trajectory of each dynamic obstacle over a future period (e.g., 2-3 seconds) based on the recorded trajectory. This trajectory can be a sequence of coordinate points, a curve composed of multiple trajectory points, or it can include motion information such as the current velocity and acceleration of each trajectory point. The specific content of the trajectory is not limited here.

[0032] In the method of this embodiment, the collision time, that is, the time it takes for the dynamic obstacle to reach the area where the car door is opened, can be calculated based on the predicted trajectory, current speed and acceleration of the dynamic obstacle.

[0033] Furthermore, the method in this embodiment also determines the prediction confidence level corresponding to each dynamic obstacle based on the obstacle's motion state (such as rate of change of velocity and rate of change of direction), historical motion trajectory, and category. Specifically, a confidence assessment model (such as a rule-based model or a lightweight neural network) can be used, taking the features of a specific dynamic obstacle as input data, and the model outputs the prediction confidence level for that obstacle. This prediction confidence level can be a value between 0 and 1. A higher value indicates higher accuracy in predicting the obstacle's trajectory, meaning the obstacle's trajectory is relatively ordered and stable. A lower value indicates lower accuracy in predicting the obstacle's trajectory, meaning the obstacle's trajectory is relatively disordered and unpredictable.

[0034] Taking prediction confidence as an example, if a dynamic obstacle is a bicycle, based on its motion state and historical trajectory, the bicycle's motion pattern is approximately a straight line at a constant speed. Therefore, the prediction confidence value for this bicycle can be 0.92. Correspondingly, based on the characteristics of a dynamic obstacle, it can be determined that it is a pedestrian hesitating at an intersection; then the prediction confidence value for this pedestrian can be 0.58. Based on the characteristics of a dynamic obstacle, it can be determined that it is an electric scooter accelerating to overtake; then the prediction confidence value for this electric scooter can be 0.43.

[0035] Step 102: Generate a risk vector based on the collision information.

[0036] In the method of this embodiment, the limit opening angle corresponding to each static obstacle, the collision time corresponding to each dynamic obstacle, and the prediction confidence are combined to obtain a multi-dimensional risk vector, which is used to characterize the risk of the car door colliding with various obstacles.

[0037] Step 103: Generate an internal intent vector based on the in-vehicle information of the target vehicle.

[0038] In this embodiment, the in-vehicle information mainly includes handlebar grip force signals, occupant posture information, and seat information. The handlebar grip force signals can be obtained from a flexible pressure sensor array (8×8 or 16×16 dot matrix, depending on requirements) deployed on the handlebar surface, collecting grip force distribution data at a preset sampling frequency (e.g., 100Hz or higher). Subsequently, the average grip force and grip force rise rate are calculated from the grip force distribution data, and the fluctuation characteristics of the grip force are analyzed to obtain the final handlebar grip force signal.

[0039] As an alternative, a single-point pressure sensor and capacitive touch sensing could be used to replace the flexible pressure sensor array to reduce costs, or a wearable surface electromyography sensor could be used to detect forearm muscle activity to improve detection accuracy.

[0040] In the method of this embodiment, grip strength fluctuation characteristics can reflect whether the act of opening the car door is tentative or eager. For example, the presence of small, short-term grip strength fluctuations indicates that the user only intends to tentatively open the car door. Meanwhile, the rate of increase in grip strength can reflect the eagerness to open the car door, and its formula is as follows: (1) In the formula, α is the rate of increase of grip force, and F is the grip force at the current time t.

[0041] As a feasible solution, user intent can be categorized and a handle intent vector generated based on the aforementioned average grip strength, grip strength rise rate, and fluctuation characteristics. Specifically, if the average grip strength is less than a preset grip strength range and the grip strength rise rate is less than a preset rate range, the handle intent vector is determined to be weak. If the average grip strength is within the preset grip strength range and the grip strength rise rate is within the preset rate range, the handle intent vector is determined to be normal. If the average grip strength is greater than the preset grip strength range and the grip strength rise rate is greater than the preset rate range, the handle intent vector is determined to be strong.

[0042] For example, when the average grip strength is low and the rate of increase in grip strength is slow with slight fluctuations, it indicates that the current user simply wants to "test the car door" and may be trying to observe the external situation. When the average grip strength is medium and the rate of increase in grip strength is stable, it indicates that the current user is opening the car door normally after confirming it is safe. However, when the average grip strength is high and the rate of increase in grip strength is steep, it indicates that the user is eager to open the car door and get out, and may not have fully observed the external situation.

[0043] In the method of this embodiment, the handle intent vector is used to characterize the intensity of the person opening the car door's intention to open the door. The content of the user intent classification is only an example. In actual work, the classification threshold, level and other parameters can be set according to the needs. The content of the user intent classification is not specifically limited here. It is sufficient to determine that the user has different degrees of intention to open the car door.

[0044] In this embodiment, the occupant posture information can come from an in-cabin camera, which can capture images of the face and upper body of each occupant. Computer vision algorithms are used for feature extraction, and head turning angle, torso twisting angle, hand position, and gaze focus area are detected and output in real time. As one feasible solution, the camera can be a driver monitoring system (DMS) camera that extends to all vehicle seats, or it can be millimeter-wave radar or ultra-wideband radar, etc.

[0045] A visual intent vector can be generated based on the posture information of drivers and passengers, primarily representing the level of preparedness of the person opening the car door for getting out of the vehicle. Specifically, the degree of observation of the target vehicle's rearview mirror / side window by the driver or passenger can be determined based on the head turning angle and the focal area of ​​their gaze. The tendency of the driver or passenger to make the action of getting out of the vehicle can be determined based on the torso twisting angle and whether the hand position is close to the door handle. Finally, a visual intent vector is generated based on the degree of observation and the trend. This visual intent vector can be represented by a value between 0 and 1. A higher value reflects a more adequate level of preparation for getting out of the vehicle, while a lower value reflects an insufficient level of preparation.

[0046] In this embodiment, the alighting intention vector can be derived from pressure distribution sensors within the seat back and cushion. Further feature extraction is performed on the collected pressure distribution data to calculate the seat's center of pressure (COP), monitoring its pressure displacement trend and velocity over time. Based on the COP, its pressure displacement trend, and pressure displacement velocity, an alighting intention vector is generated to characterize the strength of the person opening the door's intention to alight. A significant shift of the pressure center towards the door can also help confirm the user's alighting intention. The alighting intention vector can be represented as a value between 0 and 1; a higher value indicates a stronger intention to alight, while a lower value indicates a weaker intention.

[0047] As a feasible solution, seatbelt tension sensors can be used to detect changes in seatbelt tension when occupants lean forward, and the minute movements before the door opens can be analyzed through door micro-switch timing analysis. This would improve the accuracy of single-seat information recognition.

[0048] Finally, the handle intent vector, visual intent vector, and exit intent vector can be input into a lightweight neural network (such as an attention-based fusion network) for fusion. This network also incorporates the user intent classification results to generate and output an internal intent vector. This internal intent vector can be a value between 0 and 1, with a higher value indicating a stronger desire to open the car door.

[0049] Step 104: Control the opening resistance of the door according to the risk vector and the internal intent vector.

[0050] In this embodiment, the door control model receives information about the aforementioned risk vector and internal intent vector, as well as the current door damping force, to make decisions regarding door control logic. Specifically, this door control model can employ a hierarchical reinforcement learning framework, where the upper layer analyzes the risk vector based on reinforcement learning, evaluates risk patterns, and determines risk coefficients. The risk patterns can be set according to actual needs; therefore, the content and threshold of the risk patterns are not specifically limited here.

[0051] For example, a safe mode can be defined as having no dynamic obstacles and a relatively far distance from static obstacles. A warning mode can be defined as having dynamic obstacles, but if the collision time between the door and the dynamic obstacle is greater than 2 seconds. An intervention mode can be defined as having dynamic obstacles, but if the collision time is less than 1 second or if a collision with a static obstacle is imminent. A conservative mode can be defined as having dynamic obstacles, but if the prediction confidence of the dynamic obstacles is low. (This mode provides a larger safety margin compared to the intervention mode.)

[0052] As a feasible solution, the door control model can also employ fuzzy logic control, which blurs risk patterns and intentions, and outputs the results through an expert rule table. Furthermore, a system dynamics model can be constructed, and the optimal damping trajectory can be solved online based on model predictive control.

[0053] The risk coefficient characterizes the degree of control over door opening resistance. It can be a value between 0 and 1. A higher value indicates less intervention is needed to control door opening resistance, allowing manual door operation with resistance close to the basic feel. A lower value indicates a need to ensure passenger safety, requiring intervention to control door opening resistance. As a feasible solution, it can be linked to risk modes; for example, the risk coefficients in safety and alert modes are lower than those in intervention and conservative modes.

[0054] In this embodiment, the lower layer of the door control model employs a dynamic resistance calculation strategy. First, a basic damping component is determined based on the current opening angle of the door. This basic damping component increases with the opening angle, simulating the mechanical feel of opening the door, and can be represented by a curve. Then, an interactive damping component is determined based on the current risk vector and internal intention vector. This component embodies game theory and can be designed as a function proportional to the risk mode and the urgency of opening. Finally, the basic damping component and the interactive damping component are weighted and fused using a risk coefficient to obtain the opening resistance. The specific formula is as follows: F target =β F base (θ)+(1-β) F interactive (2) In the formula, F target To determine the resistance level, β is the risk coefficient, and F is the resistance level. base (θ) is the basic damping component curve, F interactive For interactive damping components.

[0055] In this embodiment, the calculated opening resistance is sent as a control command to the damping execution module. Simultaneously, the information obtained from each step of this embodiment can be synchronized to the central control screen, head-up display, or any entertainment screen in the vehicle to provide users with intuitive information. For example, when the door opens, colors can be used to indicate the current risk: green for safety, yellow for warning, and red for danger. The remaining safe distance can also be precisely displayed, such as "5cm from obstacle."

[0056] The damping execution module is responsible for accurately and quickly converting the digital commands from the decision-making layer into physical damping forces that the user can actually feel, i.e., controlling the opening resistance of the car door. This damping force can be continuously variable damping or multi-level discrete damping.

[0057] Specifically, the damping execution module receives the opening resistance, compares it with the current actual output damping force, and calculates the precise control current using a proportional-integral-derivative controller. As one feasible solution, the control current can then be applied to the magnetorheological fluid damper, changing the magnetic field strength and causing a millisecond-level change in the viscosity of the magnetorheological fluid inside the damper, thereby precisely and continuously adjusting the output damping force. Another feasible solution is to precisely change the position of the door limiter by controlling the rotation of a stepper motor, thus producing a corresponding damping effect. An electromagnetic powder clutch can also be used to achieve a high response speed. Yet another feasible solution is to directly control the door movement by driving the active door hinge directly with a motor.

[0058] In the method of this embodiment, the damping force felt by the person opening the car door is fed back to the brain through their muscle perception, influencing their next door-opening behavior (whether to continue applying force, stop, or try again). This new behavior is then captured by the internal intention perception module, thus forming a complete human-computer interaction closed loop.

[0059] In the method of this embodiment, the information in each of the above steps can be recorded to reflect the door-opening habits of a fixed person opening the door in different scenarios (such as some users' habit of "opening it a crack first to observe before opening it fully"). Through long-term learning, the door control method of this embodiment is continuously optimized to customize personalized damping curves for different users, achieving a personalized intelligent experience. For example, for cautious users, the damping intervention can be slightly later and gentler. For impatient users, the damping intervention can be earlier and slightly stronger.

[0060] As a feasible solution, the method described in this embodiment can be extended to scenarios such as trunk opening and window raising / lowering, and can be integrated with shared mobility platforms. For example, before the person opens the car door, a door opening safety prompt can be displayed simultaneously on the APP to form a double guarantee. It can also be combined with vehicle-to-everything (V2X) communication to receive information on "outside-visibility" traffic participants from roadside units, further improving safety.

[0061] As a feasible solution, the method described in this embodiment can be integrated with an autonomous driving system. When the vehicle is in autonomous driving mode, the optimal parking position can be planned in advance to reserve sufficient space for door opening. Upon arrival at the destination, the system can automatically select "right-side parking" (passenger side next to the sidewalk) to reduce the risk of door opening.

[0062] The method described in this embodiment uses multi-sensor fusion to construct the 3D contour of static obstacles and predict the trajectory of dynamic obstacles. It further improves risk identification accuracy and robustness by combining prediction confidence, reserving a larger safety margin for dynamic obstacles with random movement and high prediction difficulty, and adapting to complex road conditions across all scenarios. It can accurately predict the collision risk between the car door and dynamic obstacles to the side and rear, as well as surrounding static obstacles, eliminating scratches and injuries caused by blind spots. Compared to traditional alarm or locking solutions, it can prevent car door collisions earlier and more accurately, fundamentally improving car door safety. It employs a three-modal fusion perception of grip force, vision, and seat pressure to accurately obtain the user's intention to open the door, distinguishing between tentative opening, normal opening, and urgent opening intentions. Furthermore, it controls the door opening angle through continuous and variable real-time resistance dynamic game control, with damping force continuously and smoothly changing with the opening angle and risk level. In high-risk situations, the system prioritizes enhanced protection, while in low-risk situations, the user prioritizes maintaining a smooth feel, achieving a dynamic balance between safety and user experience. Abandoning the rigid door locking logic, it achieves "human-vehicle tactile dialogue," which not only meets the need for observation when opening the door slightly, but also allows for timely intervention in high-risk situations, solving the problems of rigid intervention and poor user experience of traditional solutions.

[0063] The method described in this embodiment requires no manual intervention and can be continuously optimized, achieving a closed-loop operation of perception-decision-execution-feedback at high frequency. It automatically completes risk identification, intent judgment, and damping adjustment, and can also record user door-opening habits to continuously optimize damping control and achieve personalized adaptive adjustment. Furthermore, the structure and algorithm of this embodiment are compatible with existing automotive hardware solutions based on mass-produced automotive hardware such as surround-view cameras, millimeter-wave radar, DMS cameras, and seat sensors, eliminating the need for numerous additional dedicated sensors. Combined with the millisecond-level response and high-precision proportional-integral control of the magnetorheological fluid damper, it can be quickly integrated into existing vehicle models. From another perspective, the tactile feedback from the door can also subtly cultivate safe door-opening habits among drivers and passengers.

[0064] Example 2: This application discloses another method for door control; please refer to [link / reference]. Figure 2 This embodiment describes the method for controlling the vehicle door.

[0065] Step 201: Scan and acquire static obstacles within 10m around the vehicle 2 seconds before the vehicle stops.

[0066] Step 202: Construct a three-dimensional contour model of the static obstacle.

[0067] Step 203: Obtain the distance between the vehicle door and the 3D contour model at each opening angle.

[0068] Step 204: Determine the maximum opening angle of the door based on the distance to the 3D contour model.

[0069] Step 205: Scan and acquire dynamic obstacles to the side and rear of the vehicle 2 seconds before the vehicle stops.

[0070] Step 206: Predict the trajectory of dynamic obstacles.

[0071] Step 207: Calculate the collision time between the dynamic obstacle and the car door based on the motion trajectory.

[0072] Step 208: Determine the prediction confidence level based on the motion state, historical motion trajectory, and category of the dynamic obstacle.

[0073] Step 209: Generate a risk vector characterizing the collision risk between the door and the obstacle based on the limit opening angle, collision time, and prediction confidence. Proceed to Step 214.

[0074] Step 210: Collect handlebar grip force signals using a flexible pressure sensor array and analyze them to obtain the handlebar intention vector.

[0075] Step 211: Collect the posture information of the driver and passengers through the in-cabin camera and analyze it to obtain the visual intent vector.

[0076] Step 212: Collect pressure distribution information in the seat back and seat cushion, and analyze it to obtain the intention vector for getting out of the vehicle.

[0077] Step 213: Fuse the handlebar intent vector, visual intent vector, and exit intent vector to generate an internal intent vector. Proceed to Step 216.

[0078] Step 214: Determine the risk coefficient based on the analysis of the risk vector.

[0079] Step 215: Determine the basic damping component curve based on the current opening angle of the car door to simulate the feel of opening the car door.

[0080] Step 216: Determine the interaction damping components based on the risk vector and the internal intent vector.

[0081] Step 217: Use the risk coefficient to weight and fuse the basic damping component and the interactive damping component to obtain the opening resistance, and apply resistance to the opening of the door dynamically.

[0082] Step 218: Convert the risk of opening the door mapped by the opening resistance into text, images, or voice and display it on the vehicle's display screen.

[0083] Example 3: This application discloses another method for controlling vehicle doors. The method described in this example is for the process of a person tentatively opening a vehicle door in a high-risk situation.

[0084] At time t0, it is detected that person A lightly touches the handle and applies slight pressure to test it. The external risk is: a high-speed electric vehicle is approaching, with a collision time of 1.8 seconds and a confidence level of 0.9. The internal intent is: low average grip force and a gradual rate of increase in grip force. The handle intent vector is determined to be at the level of probing, with a risk coefficient of 0.1, indicating strong intervention is required. Therefore, the door opening resistance is set to strong (e.g., 8 N·m). Person A will feel the door is "unmoving," as if it is firmly sucked in, and will be given both voice and visual prompts: "Vehicle approaching from behind, please do not open the door."

[0085] At time t1, person A who opens the car door immediately stops trying to open it. After the electric vehicle passes, the risk factor changes to 0.8. At this point, the opening resistance can be restored to the basic mechanical feel, and the car door can be opened normally again.

[0086] Example 4: This application discloses another method for controlling vehicle doors. The method described in this example is for the process of a person eagerly opening a vehicle door during a medium-risk situation.

[0087] At time t0, it is detected that person B is urgently pulling the door open with considerable force. The external risk at this time is: a slow-moving bicycle approaching, with a collision time of 2.5 seconds and a confidence level of 0.85. The internal intent is: a relatively high average grip force with a steep rate of increase, determining the handle intent vector as strong and the risk coefficient as 0.4. Therefore, the door opening resistance is set to a base damping plus an additional damping that increases exponentially with the opening angle. When person B opens the door at a small angle (approximately 10°), they will feel that "it can be opened, but the door becomes increasingly heavy."

[0088] At time t1, person B, who is opening the car door, feels increased resistance and slows down the opening speed. The bicycle approaches, and the collision time is 1.5 seconds. The door damping continues to increase, reaching its peak when the door opening angle is 15°. At this point, person B will intuitively perceive that "the door cannot be opened any further" and stop opening the door. At time t2, the bicycle passes, the risk is eliminated, the door opening resistance returns to its basic mechanical feel, and person B can open the door and get out normally.

[0089] Example 5: This application discloses another method for controlling a car door. The method described in this example is for the process of opening a car door when there is a static obstacle.

[0090] Person C opens the car door normally at time t0. The external risks to the vehicle at this time are: a pillar is present on the left side; the distance to the pillar is 5cm when the door is open at 20°, 2cm when the door is open at 22°, and 0cm (collision) when the door is open at 23°. Therefore, the risk factor is determined to be 0.7, and additional damping is introduced starting at an opening angle of 18°, reaching its peak at an opening angle of 22°.

[0091] As person C opens the car door, they will feel significant resistance when the door is 20° open, and will receive a voice or image prompt saying "3cm remaining to the obstacle".

[0092] At time t1, person C attempts to open the car door with slight force, but feels that it cannot be opened further when the door opens to an angle of 22°. At this point, the door is restricted to a safe range, and person C receives precise distance feedback to understand the limit of the opening angle, thus maximizing the use of the door opening space as needed.

[0093] Based on the door control method disclosed in the above embodiments, this embodiment discloses a door control device. Please refer to... Figure 3 The door control device includes: a prediction unit 301, a risk generation unit 302, an intent generation unit 303, and a door control unit 304; The prediction unit 301 is used to predict collision information when the door of the target vehicle collides with an obstacle; the obstacle is located within a preset range of the target vehicle and includes static obstacles and dynamic obstacles; the collision information includes the limit opening angle when the door collides with the static obstacle, the collision time between the door and the dynamic obstacle, and the prediction confidence level corresponding to the dynamic obstacle; The risk generation unit 302 is used to generate a risk vector based on the collision information; the risk vector is used to characterize the risk of the door colliding with the obstacle. The intent generation unit 303 is used to generate an internal intent vector based on the in-vehicle information of the target vehicle; the in-vehicle information includes handle grip force signal, driver and passenger posture information, and seat information; the internal intent vector is used to represent the intention of the person opening the door to open the door. The door control unit 304 is used to control the opening resistance of the door according to the risk vector and the internal intent vector.

[0094] Optionally, the prediction unit 301 includes: An angle prediction subunit is used to predict the limit opening angle based on the three-dimensional contour model of the static obstacle. A trajectory prediction subunit is used to predict the motion trajectory of the dynamic obstacle; A time prediction subunit is used to calculate the collision time based on the motion trajectory; The confidence prediction subunit is used to determine the prediction confidence based on the motion state, historical motion trajectory and category of the dynamic obstacle.

[0095] Optionally, the angle prediction subunit includes: A scanning subunit is used to scan within a preset range when the target vehicle is stopped using an onboard sensing device to obtain scanning data; the onboard sensing device includes a camera and an ultrasonic radar. A point cloud construction subunit is used to construct a point cloud map containing the static obstacles based on the scanned data; The model building subunit is used to build the three-dimensional contour model of the static obstacles in the point cloud map; The distance calculation subunit is used to calculate the distance between the car door and the three-dimensional contour model at each opening angle, based on the configuration information of the car door. The limit prediction subunit is used to predict the limit opening angle based on the distance.

[0096] Optionally, the confidence prediction subunit includes: The confidence determination subunit is used to input the motion state, the historical motion trajectory, and the category into a pre-trained confidence evaluation model to obtain the predicted confidence output by the confidence evaluation model; the motion state includes the rate of change of speed and the rate of change of direction; the category includes pedestrians, motor vehicles, and non-motor vehicles.

[0097] Optionally, the intent generation unit 303 includes: A handle intention generation subunit is used to generate a handle intention vector based on the handle grip force signal; the handle intention vector is used to characterize the intensity of the person opening the car door's intention to open the door. A visual intent generation subunit is used to generate a visual intent vector based on the posture information of the driver and passengers; the visual intent vector is used to characterize the degree of preparation of the person opening the car door for getting out of the car. The feature extraction subunit is used to extract features from the collected seat information to obtain the seat's pressure center, pressure displacement trend, and pressure displacement velocity. The vehicle exit intention generation subunit is used to generate the vehicle exit intention vector based on the pressure center, the pressure displacement trend, and the pressure displacement velocity; the vehicle exit intention vector is used to characterize the intensity of the person opening the car door's intention to get out of the vehicle; The fusion subunit is used to fuse the handle intent vector, the visual intent vector, and the alighting intent vector to generate the internal intent vector.

[0098] Optionally, the handle is intended to generate a sub-unit, including: The grip force extraction subunit is used to extract features from the collected handlebar grip force signal to obtain the average grip force and grip force rise rate. The weak determination subunit is used to determine that the handle intention vector is weak when the average grip force is less than a preset grip force range and the grip force rise rate is less than a preset rate range. The normal determination subunit is used to determine that the handle intention vector is normal when the average grip force is within the preset grip force range and the grip force rise rate is within the preset rate range. A strongly determined subunit is used to determine the handle intent vector as strong when the average grip force is greater than the preset grip force range and the grip force rise rate is greater than the preset rate range.

[0099] Optionally, the visual intent generation subunit includes: The posture extraction subunit is used to extract features from the collected posture information of the driver and passengers, and obtain the head turning angle, torso twisting angle, hand position and gaze focus area. The observation and determination subunit is used to determine the degree of observation of the driver / passenger of the target vehicle's rearview mirror / side window based on the head turning angle and the focal area of ​​the line of sight. The trend determination subunit is used to determine the trend of the driver / passenger making an exit action based on the torso twist angle and the hand position; A vector generation subunit is used to generate the visual intent vector based on the observation level and the trend.

[0100] Optionally, the door control unit 304 includes: A coefficient determination subunit is used to determine a risk coefficient based on the risk vector; the risk coefficient is used to characterize the degree to which the opening resistance is controlled. A basic damping determination subunit is used to determine the basic damping component based on the current opening angle of the door; the basic damping component is used to simulate the feel of opening the door. An interactive damping determination subunit is used to determine the interactive damping component based on the risk vector and the internal intent vector. The weighted fusion subunit is used to weight and fuse the basic damping component and the interactive damping component using the risk coefficient to obtain the opening resistance.

[0101] Based on the above-mentioned door control method, this application also discloses a vehicle, including: a processor, a memory, and a system bus; The processor and the memory are connected via the system bus; The memory is used to store a program, the program including instructions that, when executed by the processor, cause the processor to perform the steps of any of the methods described above.

[0102] The embodiments in this specification are described in a progressive manner. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant details can be found in the method section.

[0103] It should also be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0104] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented directly by hardware, a software module executed by a processor, or a combination of both. The software module can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.

[0105] The features described in the embodiments of this specification can be substituted for or combined with each other, so that those skilled in the art can implement or use this application.

[0106] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for controlling vehicle doors, characterized in that, include: Predict collision information when the target vehicle's door collides with an obstacle; The obstacle is located within a preset range of the target vehicle and includes static obstacles and dynamic obstacles; the collision information includes the limit opening angle when the door collides with the static obstacle, the collision time between the door and the dynamic obstacle, and the prediction confidence level corresponding to the dynamic obstacle; A risk vector is generated based on the collision information; The risk vector is used to characterize the risk of the door colliding with an obstacle; An internal intent vector is generated based on the in-vehicle information of the target vehicle. The in-vehicle information includes handlebar grip force signals, driver and passenger posture information, and seat information; The internal intent vector is used to characterize the intention of the person opening the car door; The opening resistance of the vehicle door is controlled based on the risk vector and the internal intent vector.

2. The method according to claim 1, characterized in that, The collision information predicted when the target vehicle's door collides with an obstacle includes: Based on the three-dimensional contour model of the static obstacle, the maximum opening angle is predicted; Predict the trajectory of the dynamic obstacle; The collision time is calculated based on the motion trajectory; The prediction confidence level is determined based on the motion state, historical motion trajectory, and category of the dynamic obstacle.

3. The method according to claim 2, characterized in that, The prediction of the ultimate opening angle based on the three-dimensional contour model of the static obstacle includes: The vehicle-mounted sensing device scans the preset range when the target vehicle stops, and obtains scanning data; the vehicle-mounted sensing device includes a camera and an ultrasonic radar. A point cloud map containing the static obstacles is constructed based on the scan data; The three-dimensional contour model of the static obstacles in the point cloud map is established; Based on the configuration information of the car door, the distance between the car door and the three-dimensional contour model at each opening angle is calculated; The maximum opening angle is predicted based on the distance.

4. The method according to claim 2, characterized in that, The determination of the prediction confidence level based on the motion state, historical trajectory, and category of the dynamic obstacle includes: The motion state, the historical motion trajectory, and the category are input into a pre-trained confidence evaluation model to obtain the predicted confidence level output by the confidence evaluation model; the motion state includes the rate of change of speed and the rate of change of direction; the category includes pedestrians, motor vehicles, and non-motor vehicles.

5. The method according to claim 1, characterized in that, The generation of an internal intent vector based on the in-vehicle information of the target vehicle includes: A handle intention vector is generated based on the handle grip force signal; the handle intention vector is used to characterize the intensity of the person opening the car door's intention to open the door. A visual intent vector is generated based on the posture information of the driver and passengers; the visual intent vector is used to characterize the degree of readiness of the person opening the car door to get out of the car. Feature extraction is performed on the collected seat information to obtain the seat's pressure center, pressure displacement trend, and pressure displacement velocity. Based on the pressure center, the pressure displacement trend, and the pressure displacement velocity, the disembarkation intention vector is generated; the disembarkation intention vector is used to characterize the intensity of the person opening the car door's intention to get out of the car. The handle intention vector, the visual intention vector, and the alighting intention vector are fused to generate the internal intention vector.

6. The method according to claim 5, characterized in that, The generation of the handlebar intention vector based on the handlebar grip force signal includes: Feature extraction is performed on the collected handlebar grip force signal to obtain the average grip force and grip force rise rate; If the average grip force is less than a preset grip force range and the grip force rise rate is less than a preset rate range, the handle intention vector is determined to be weak. If the average grip force is within the preset grip force range and the grip force increase rate is within the preset rate range, the handle intention vector is determined to be normal. If the average grip force is greater than the preset grip force range and the grip force increase rate is greater than the preset rate range, the handle intention vector is determined to be strong.

7. The method according to claim 5, characterized in that, The generation of visual intent vectors based on the occupant posture information includes: Feature extraction is performed on the collected posture information of the driver and passengers to obtain the head turning angle, torso twisting angle, hand position, and gaze focus area; Based on the head turning angle and the focal area of ​​vision, determine the degree of observation of the target vehicle's rearview mirror / side window by the driver / passenger. Based on the torso twist angle and the hand position, determine the tendency of the driver / passenger to make an exit action; The visual intent vector is generated based on the degree of observation and the trend.

8. The method according to claim 1, characterized in that, The step of controlling the opening resistance of the vehicle door based on the risk vector and the internal intent vector includes: A risk coefficient is determined based on the risk vector; the risk coefficient is used to characterize the degree to which the opening resistance is controlled. The basic damping component is determined based on the current opening angle of the door; the basic damping component is used to simulate the feel of opening the door. The interaction damping component is determined based on the risk vector and the internal intent vector. The opening resistance is obtained by weighting and fusing the basic damping component and the interactive damping component using the risk coefficient.

9. A door control device, characterized in that, include: Prediction unit, risk generation unit, intent generation unit, and door control unit; The prediction unit is used to predict collision information when the door of the target vehicle collides with an obstacle. The obstacle is located within a preset range of the target vehicle and includes static obstacles and dynamic obstacles; the collision information includes the limit opening angle when the door collides with the static obstacle, the collision time between the door and the dynamic obstacle, and the prediction confidence level corresponding to the dynamic obstacle; The risk generation unit is used to generate a risk vector based on the collision information; The risk vector is used to characterize the risk of the door colliding with an obstacle; The intent generation unit is used to generate an internal intent vector based on the in-vehicle information of the target vehicle. The in-vehicle information includes handlebar grip force signals, driver and passenger posture information, and seat information; The internal intent vector is used to characterize the intention of the person opening the car door; The door control unit is used to control the opening resistance of the door based on the risk vector and the internal intent vector.

10. A vehicle, characterized in that, include: Processor, memory, system bus; The processor and the memory are connected via the system bus; The memory is used to store a program, the program including instructions that, when executed by the processor, cause the processor to perform the steps of the method according to any one of claims 1-8.

Images