A proactive warning method and system for door opening risk based on differentiated hierarchical control

By using a multimodal sensing unit and a hierarchical risk assessment model, a three-dimensional risk representation is generated and differentiated warning instructions are generated. This solves the problem of insufficient multi-dimensional risk identification when the car door is opened, and improves the comprehensiveness and scenario adaptability of the car door opening risk warning, especially in the scenario of occupants resting after the vehicle is turned off.

CN121084286BActive Publication Date: 2026-11-10CHERY NEW ENERGY AUTOMOBILE TECH CO LTD
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202511200255.7
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2025-08-26
Publication Date
2026-11-10
Estimated Expiration
2045-08-26

AI Technical Summary

Technical Problem

Existing technologies lack the ability to simultaneously collect multi-dimensional environmental information around car doors, making it impossible to differentiate and respond to various risks. This results in insufficient comprehensive safety warnings when car doors are opened, especially in the scenario where passengers are resting after the vehicle is turned off.

Method used

The system monitors the surrounding environment of the car door in real time through a multimodal sensing unit, generates a three-dimensional risk representation, dynamically identifies the risk level using a graded risk judgment model, and generates differentiated warning instructions based on the level, including ambient lighting, sound field speakers and haptic feedback, and performs incremental risk calculation in rest scenarios.

Benefits of technology

It enables comprehensive identification and precise classification and response to multiple risks, improves the comprehensiveness and pertinence of door opening risk warning, fills the safety protection gap in special scenarios, and enhances the safety of people inside the vehicle.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN121084286B_ABST
    Figure CN121084286B_ABST
Patent Text Reader

Abstract

This invention proposes a proactive warning method and system for vehicle door opening risks based on differentiated hierarchical control, belonging to the field of new energy vehicle safety technology. It includes: real-time monitoring of vehicle status; activation of a multimodal sensing unit when the vehicle is parked and at least one door opening signal is triggered; synchronous acquisition of static obstacle height information, dynamic living target behavior trajectory, and spatial distribution of low obstacles in the surrounding environment of the door through the multimodal sensing unit to generate a three-dimensional risk representation; dynamic identification of risk levels and generation of differentiated warning commands based on the three-dimensional risk representation using a hierarchical risk judgment model; and automatic reset of the proactive warning system when a door closing signal is detected and the risk is eliminated. This invention, through a comprehensive warning mechanism, can significantly improve the warning effect of vehicle door opening risks, thereby providing better protection for the safety of occupants.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention belongs to the field of new energy vehicle safety technology, and in particular relates to a method and system for proactive early warning of door opening risks based on differentiated hierarchical control. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] With the rapid popularization of new energy vehicles, composite materials, due to their advantages such as lightweight, flexible design, and corrosion resistance, are widely used in car door manufacturing, driving car door designs towards more complex and larger sizes. However, safety issues during the car door opening process are becoming increasingly prominent. Drivers and passengers are easily distracted by the surrounding environment when opening the door and may overlook potential risks, leading to accidents such as door damage or personal injury.

[0004] While some technical solutions related to vehicle door safety exist in the existing technology, these solutions generally suffer from some intractable technical problems, such as: (1) Existing technologies mostly adopt a single early warning or protection method, which lacks the ability to collect multi-dimensional environmental information simultaneously on static high obstacles (such as vehicles, pillars, and walls of equal or higher height than the vehicle), low obstacles (such as curbs and guardrails) and dynamic living targets (such as children playing and pets) around the vehicle door. Furthermore, it has not established a hierarchical judgment logic for different types of risks, resulting in the inability to comprehensively identify and differentiate various risks, and the comprehensiveness of early warning is insufficient.

[0005] (2) The existing technology does not take into account the special scenario of passengers resting in the vehicle after the engine is turned off. When the door is closed and then reopened, it cannot provide effective risk warning for the door opening needs in such scenarios, resulting in a gap in safety protection in special use scenarios and insufficient scenario adaptability. Summary of the Invention

[0006] To overcome the shortcomings of the prior art, this invention provides a proactive warning method and system for door opening risks based on differentiated hierarchical control. Through a comprehensive warning mechanism, the warning effect of door opening risks can be significantly improved, thereby providing better protection for the safety of people inside the vehicle.

[0007] To achieve the above objectives, one or more embodiments of the present invention provide the following technical solutions: The first aspect of this invention provides a proactive warning method for door opening risk based on differentiated hierarchical control.

[0008] The proactive warning method for door opening risk based on differentiated hierarchical control includes: Real-time monitoring of vehicle status; when the vehicle is detected to be parked and at least one door is triggered, the multimodal sensing unit is activated. The multimodal sensing unit synchronously collects static obstacle height information, dynamic live target behavior trajectory and low obstacle spatial distribution information around the vehicle door, and generates a three-dimensional risk representation. Based on the obtained three-dimensional risk representation, a graded risk assessment model is used to dynamically identify the risk level. Differentiated warning commands are generated based on the risk level: for low-risk scenarios, the ambient lights are controlled to output monochromatic steady-state light; for medium- and high-risk scenarios, the ambient lights are driven to execute a multi-frequency flashing sequence and the directional sound field speakers are activated. When a door closing signal is detected and the risk is eliminated, the active warning system is automatically reset.

[0009] Furthermore, the activation of the multimodal sensing unit includes: first, waking up the millimeter-wave radar array corresponding to the target door through the body controller to construct a polar coordinate monitoring network with the door as the origin; and simultaneously activating the wide-angle fisheye camera to perform dynamic video analysis on the fan-shaped area of ​​the door opening.

[0010] Furthermore, the construction of the graded risk assessment model includes: first, establishing a static obstacle height-distance mapping table, i.e., when the obstacle height is higher than the vehicle height and the distance is less than a first threshold, it is judged as high risk; then, constructing a low obstacle feature library, i.e., identifying the contour features of curbs and guardrails through a machine learning model, triggering a scratch warning when the distance is less than a second threshold; finally, deploying a live target behavior prediction model, i.e., calculating the motion vector of a live target in real time based on the YOLOv7 framework and predicting the probability of entering a dangerous area.

[0011] Furthermore, the execution of the differentiated warning command includes: pre-setting multi-level color gamut encoding in the LED light strip embedded in the car door, with green indicating safety, yellow indicating warning, and red indicating high risk; generating an adjustable focused sound beam through a directional speaker to concentrate sound energy in the direction of risk; and superimposing a pulse vibration warning from a tactile feedback module when a live target is detected approaching rapidly.

[0012] Furthermore, the proactive warning method for door opening risk based on differentiated hierarchical control also includes a rest scenario re-triggering mechanism. Specifically, if the door experiences a closing and reopening event after the vehicle is turned off, and the seat sensor detects the continued presence of the occupant, the environmental data collected before hibernation is reused, and incremental risk calculation is performed only on the newly added variables.

[0013] Furthermore, the incremental risk calculation includes: detecting the vehicle position offset through the on-board IMU sensor, re-initializing the sensing unit when the offset is greater than a set value; marking the motion state of static obstacles, and performing continuous tracking only on unmarked targets.

[0014] The second aspect of the present invention provides a proactive warning system for door opening risks based on differentiated hierarchical control.

[0015] A door opening risk proactive warning system based on differentiated hierarchical control includes: The multimodal sensing module includes a millimeter-wave radar array, a wide-angle fisheye camera, and an ultrasonic sensor arranged around the door. The hierarchical decision-making module is used to execute the hierarchical risk assessment model and output risk level codes; The dynamic response module includes a programmable LED ambient lighting strip, a beamforming speaker, and a haptic feedback unit; The scenario adaptation module is used to realize incremental risk calculation and cloud-based collaborative optimization for rest scenarios; The vehicle body controller is used to coordinate the data flow and command transmission of the multimodal sensing module, hierarchical decision-making module, dynamic response module, and scene adaptation module.

[0016] Furthermore, the deployment method of the millimeter-wave radar array includes: installing three radar nodes embedded in the vehicle door, respectively covering monitoring sectors at three different angles: 0-30°, 30-60°, and 60-90°; each radar node uses a TDMA mechanism to transmit frequency-modulated continuous waves in a time-division manner to avoid signal interference. Furthermore, the dynamic response module also includes an AR-HUD projection unit and a piezoelectric vibrating pad built into the seat; wherein, the AR-HUD projection unit is used to overlay and display obstacle location indicators on the windshield; and the piezoelectric vibrating pad built into the seat is used to generate a tactile coding sequence that is positively correlated with the risk level.

[0017] Furthermore, the scene adaptation module includes a vehicle micro-motion sensing unit and an environmental change comparison unit; wherein, the vehicle micro-motion sensing unit is used to detect vehicle displacement using a MEMS accelerometer; and the environmental change comparison unit is used to identify scene differences through a feature point matching algorithm.

[0018] The above one or more technical solutions have the following beneficial effects: (1) This invention generates a three-dimensional risk representation by activating a multimodal sensing unit to synchronously collect information on the height of static obstacles around the vehicle door, the dynamic behavior trajectory of a living target, and the spatial distribution of low obstacles. Based on a graded risk assessment model, the risk level is dynamically identified, and differentiated warning instructions are generated according to the risk level (such as monochromatic steady-state light in low-risk scenarios, and a combination of multi-frequency flashing sequences and directional sound field speakers in medium- and high-risk scenarios). This design achieves comprehensive identification and precise graded response to multiple risk types, effectively improving the comprehensiveness and pertinence of the vehicle door opening risk warning.

[0019] (2) The rest scenario re-triggering mechanism in this invention allows for the reuse of environmental data collected before hibernation when the vehicle door experiences a closing and reopening event after the vehicle is turned off and the seat sensor detects the continued presence of the occupant. This mechanism enables rapid response to the door opening requirements in rest scenarios, reducing redundant operations of repeated initialization of the sensing unit while ensuring the accuracy of the warning. It fills the gap in safety protection in special scenarios and significantly improves the scenario adaptability of the patented technology.

[0020] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0021] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0022] Figure 1 This is a flowchart of the proactive warning method for door opening risk based on differentiated hierarchical control in Embodiment 1 of the present invention.

[0023] Figure 2 This is a modular structure diagram of the door opening risk active warning system based on differentiated hierarchical control in Embodiment 2 of the present invention.

[0024] Figure 3 This is a modular structure diagram of the dynamic response module in Embodiment 2 of the present invention.

[0025] Figure 4 This is a modular structure diagram of the scene adaptation module in Embodiment 2 of the present invention. Detailed Implementation

[0026] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0027] It should be noted that the terminology used herein is for the purpose of describing particular implementations only and is not intended to limit the exemplary implementations of the present invention.

[0028] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0029] Example 1 This embodiment discloses a proactive warning method for door opening risk based on differentiated hierarchical control.

[0030] like Figure 1 As shown, the active warning method for door opening risk based on differentiated hierarchical control includes: Step S1: Monitor vehicle status in real time. When the vehicle is detected to be parked and at least one door opening signal is triggered, activate the multimodal sensing unit. Step S2: Synchronously collect static obstacle height information, dynamic live target behavior trajectory and low obstacle spatial distribution information of the surrounding environment of the vehicle door through the multimodal sensing unit to generate a three-dimensional risk representation; Step S3: Based on the obtained three-dimensional risk representation, a graded risk assessment model is used to dynamically identify the risk level; Step S4: Generate differentiated warning instructions based on the risk level, namely: for low-risk scenarios, control the vehicle ambient lights to output monochromatic steady-state light; for medium- and high-risk scenarios, drive the ambient lights to execute a multi-frequency flashing sequence and activate the directional sound field speakers. Step S5: When a door closing signal is detected and the risk is eliminated, the active warning system is automatically reset.

[0031] Based on the above process, this invention, through a comprehensive early warning mechanism, can significantly improve the early warning effect of vehicle door opening risks, thereby providing better protection for the safety of occupants. To facilitate understanding of the technical solution of this invention, the specific implementation methods of this invention will be further explained and described below.

[0032] In step S1, the vehicle status is monitored in real time. When the vehicle is detected to be parked and at least one door opening signal is triggered, the multimodal sensing unit is activated. This can be achieved through the following methods: 1) The millimeter-wave radar array corresponding to the target door is activated by the body controller to build a polar coordinate monitoring network with the door as the origin.

[0033] After receiving a door opening signal (such as a level transition signal from a door control switch), the body controller sends a wake-up command (12V wake-up voltage, lasting 200ms) to the millimeter-wave radar array of the target door via the CAN bus. This radar array contains three embedded radar nodes (using a 77GHz frequency-modulated continuous wave radar), installed at the upper, middle, and lower edges of the door, respectively, covering... , , Horizontal monitoring sectors.

[0034] The polar coordinate network takes the center of the car door handle as the origin and the radial direction ( The distance between the obstacle and the car door is the straight-line distance (detection range 0.3-5m), and the angular distance is ( The angle between the obstacle and the direction in which the door opens is ( ). Each radar node outputs a set of point cloud data (including...) every 10ms. , (and reflection intensity). Each radar node adopts the Time Division Multiple Access (TDMA) mechanism, according to... node, node, The nodes transmit signals sequentially in a time-division manner (each radar node transmits for 10ms with a cycle of 30ms) to avoid interference between signals on the same frequency.

[0035] 2) Simultaneously activate the wide-angle fisheye camera to perform dynamic video analysis on the fan-shaped area where the car door is open.

[0036] The body controller synchronously sends a start signal (trigger current 500mA) to the wide-angle fisheye camera integrated in the door rearview mirror. The camera parameters are configured as follows: field of view. (cover The fan-shaped area and surrounding area of ​​the car door opening Redundancy range), resolution 1920×1080, frame rate 30fps, lens distortion coefficient .

[0037] The dynamic video analysis process is as follows: 1) Real-time acquisition of video frames and distortion correction via the ISP module (restoring linear features based on a preset fisheye lens intrinsic parameter matrix); 2) Setting the ROI (Region of Interest) to the opening of the car door. 3) Extract edge features (using the Canny operator) and color histograms from each frame of the image for subsequent obstacle contour matching; 4) Perform optical flow analysis on consecutive frames and mark moving regions (regions with pixel displacement > 3 pixels / frame are judged as dynamic regions).

[0038] In step S2, a three-dimensional risk representation is generated by synchronously collecting static obstacle height information, dynamic live target behavior trajectory, and spatial distribution of low obstacles in the environment surrounding the vehicle door using a multimodal sensing unit. This can be achieved through the following methods: 1) Time synchronization. Based on the vehicle-mounted high-precision clock (error... The radar point cloud (timestamp accurate to ms) and video frames (timestamps of each frame aligned to the acquisition time) are timestamp matched to ensure that the data corresponds to the environmental state at the same time.

[0039] 2) Spatial registration. Using a pre-calibrated transformation matrix (obtained offline), the radar polar coordinates (…) are registered. , Convert to camera pixel coordinates ( , Establish the spatial mapping relationship between point clouds and images.

[0040] 3) Multi-dimensional data association.

[0041] Static obstacle height information: The radar point cloud calculates the obstacle height (highest layer Z value - lowest layer Z value) by vertically layering (Z-axis, 0.1m interval), and verifies the obstacle type by combining the outline of the corresponding area in the video (such as the straight edge of a wall).

[0042] Dynamic live target behavior trajectory: Live targets (people, pets, etc.) are located in video frames using target detection boxes (YOLOv7 output), and the movement points in the radar point cloud are correlated (displacement over 3 consecutive frames). ), generate trajectory (including time, location ( , (Velocity vector).

[0043] Spatial distribution of low-lying obstacles: screening for near-range obstacles in radar point clouds ( ), low altitude ( The point set corresponds to the low contours within the ROI in the video (such as the gray-scale abrupt edge of a curb), and a distribution heatmap is generated (the higher the density, the higher the risk).

[0044] 4) Generation of 3D Risk Representation: Construct a 3D matrix, where each element is a risk quantification value (0-100, with higher values ​​indicating higher risk); the 3D matrix is ​​as follows: ; ; .

[0045] In step S3, based on the obtained three-dimensional risk representation, a graded risk assessment model is used to dynamically identify the risk level.

[0046] The construction of the graded risk assessment model includes: First, establishing a static obstacle height-distance mapping table, i.e., when the obstacle height is higher than the vehicle height and the distance is less than a first threshold, it is judged as high risk; then, constructing a low obstacle feature library, i.e., using a machine learning model to identify the contour features of curbs and guardrails, triggering a scrape warning when the distance is less than a second threshold; finally, deploying a live target behavior prediction model, i.e., calculating the motion vector of live targets in real time based on the YOLOv7 framework and predicting the probability of entering the danger zone. Specifically, this can be achieved through the following methods: 1) Establish a height-distance mapping table: When the height of an obstacle is higher than the height of the vehicle and the distance is less than the first threshold, it is judged as high risk.

[0047] The vehicle height H is preset (e.g., 1.5m, which can be read from the vehicle ECU). The first threshold is set to 1.5m (based on the safe distance under the maximum opening angle of the door). The mapping table is divided into two dimensions: "height-distance", as shown in Table 1.

[0048] Table 1 Height and Distance Mapping Table

[0049] The height and distance of static obstacles are entered into a mapping table in real time to match the corresponding risk level. For example, when a wall with a height of 1.6m is detected ( And the distance is 1.2m. Obstacles exceeding 1.5m are directly classified as high-risk. The obstacle height and distance are recalculated every 500ms, and the mapping table matching results are updated to ensure real-time risk level accuracy.

[0050] 2) Construct a feature library of low-lying obstacles: Identify the outline features of curbs and guardrails through machine learning models, and trigger a scratch warning when the distance is less than the second threshold.

[0051] Collect at least 100,000 samples (including images and radar data of curbs (0.1-0.3m high, 0.2-0.5m wide) and traffic barriers (0.3-0.6m high, with rectangular and circular cross-sections) under different lighting and weather conditions), and extract features: A. Image features: edge gradient direction (the vertical edge ratio of the curbstone is >60%), texture entropy (the entropy value of the rough surface is >5), color channel mean (the curbstone is mostly gray, and the mean value of the B channel is 100-150). B. Radar characteristics: Reflection intensity (concrete material reflection intensity > -60dBm), point cloud distribution density (per 0.1... Point cloud count > 5.

[0052] A CNN (such as ResNet-18) is used to train the classifier. The input is the fused image-radar features, and the output is "curbstone", "barrier", and "other" categories, ensuring an accuracy of ≥95%. The second threshold is set to 0.5m (the minimum safe distance between the bottom of the car door and the obstacle). When the model identifies a low obstacle and the radar detection distance is <0.5m, a scratch warning is triggered (the risk level is marked as medium risk).

[0053] 3) Deploy a live target behavior prediction model: Calculate the motion vector of the live target in real time based on the YOLOv7 framework and predict the probability of entering the danger zone.

[0054] YOLOv7 configuration: Input video frames (640×640), output bounding boxes (x, y, w, h) and confidence scores (≥0.8 for valid targets) of live targets (humans, cats, dogs, etc.), inference time per frame ≤20ms. Linear fitting is performed on the center coordinates (x_center, y_center) of the target bounding boxes for 5 consecutive frames to calculate the velocity vector, and this is combined with the radial velocity from the radar point cloud. calibration( , and deviation (Valid at the time). The velocity vector can be calculated as follows: ; ; ; in, This represents the velocity component of the live target in the horizontal direction of the image coordinate system, measured in pixels per second. The x-axis is along the horizontal direction of the image. The difference in x-coordinates between the centers of the target bounding boxes in two consecutive frames; This represents the velocity component of the live target in the vertical direction of the image coordinate system, measured in pixels per second. The y-axis is along the vertical direction of the image. The difference in y-coordinates between the centers of the target bounding boxes in two consecutive frames; This indicates the time interval between two adjacent video frames, in seconds. Since the frame rate of a wide-angle fisheye camera is 30fps, ... That is, the time difference between two adjacent frames is approximately 0.033 seconds.

[0055] Open the car door A fan-shaped area, 0-2m from the vehicle door (maximum opening radius), is marked as a danger zone. Based on this, a Kalman filter is used to predict the target's position sequence within the next 1 second. The probability of entering a location is the percentage of frames whose position falls within the danger zone. 70% is considered high risk; 30% probability 70% is considered medium risk; probability 30% is considered low risk.

[0056] In step S4, differentiated warning instructions are generated based on the risk level: for low-risk scenarios, the vehicle ambient lights are controlled to output monochromatic steady-state light; for medium- and high-risk scenarios, the ambient lights are driven to execute a multi-frequency flashing sequence and the directional sound field speakers are activated.

[0057] The execution of differentiated warning commands includes: pre-setting multi-level color gamut encoding in the LED light strip embedded in the car door, with green indicating safety, yellow indicating warning, and red indicating high risk; generating an adjustable focused sound beam through a directional speaker to concentrate sound energy in the direction of risk; and superimposing a pulse vibration warning from a tactile feedback module when a living target is detected approaching rapidly. Specifically, this includes: 1) LED light strip coding: A 1.2m long LED light strip is embedded in the door, with preset color gamut coding: Safety (risk value 0-9): Green (wavelength 520nm), constant light (brightness 50) ); Warning (Medium risk, 40-79): Yellow (wavelength 580nm), flicker frequency 2Hz (on 0.25s / off 0.25s), brightness 80 ; High risk (80-100): Red (wavelength 620nm), flicker frequency 5Hz (on 0.1s / off 0.1s), brightness 100 ; The light strip lights up in segments, corresponding to the direction of risk (e.g., If a risk is detected in an area, illuminate the corresponding 1 / 3 of the light strip.

[0058] 2) Directional speaker control: The speaker (using beamforming technology) is mounted on the door trim panel, operating at a frequency of 2-4kHz (the frequency range sensitive to human hearing), with a beamwidth of... According to the risk direction ( Adjust the beam angle (e.g., value) Then the beam center is aligned The sound pressure level is adjusted according to the risk level: 65 dB for medium risk and 85 dB for high risk, with the sound beam focused in the direction of the risk. (Deviation), to avoid interfering with other directions.

[0059] 3) Tactile feedback superposition: When the radial component of the velocity vector of the living target... (Upon rapid approach) the piezoelectric vibrator built into the door armrest is activated: Medium risk: Vibration frequency 20Hz, pulse interval 0.5s (vibration 0.2s / stop 0.3s); High risk: Vibration frequency 50Hz, continuous vibration (until the risk is eliminated), amplitude increases with approach speed (speed...). (The amplitude doubles).

[0060] Furthermore, the proactive early warning method provided by this invention also includes a rest scenario re-triggering mechanism, that is: if the vehicle door experiences a closing and reopening event after the vehicle is turned off, and the seat sensor detects the continued presence of the occupant, then the environmental data collected before hibernation is reused, and incremental risk calculation is performed only on newly added variables. Specifically, this includes: 1) Trigger condition detection.

[0061] Vehicle ignition off signal (IGN mode power off, lasting >5 seconds); Door status sequence: Closed (door control switch normally closed > 30s) Open (door switch off); The seat sensor (pressure sensor) detected a pressure value >50N (corresponding to an adult's weight), and this pressure remained constant. (To eliminate accidental touches).

[0062] 2) Data reuse: Before the system goes into hibernation (when the doors are closed), the environmental data (static obstacle positions, low obstacle distribution, areas without dynamic targets) is encrypted and stored in the local cache (capacity 10MB, storage time 30min). When the system is restarted, the data is directly called as the baseline.

[0063] 3) New variable recognition: Feature point matching between real-time camera frames and cached images (SIFT algorithm, matching degree). The area is the difference zone), combined with the newly added radar point cloud (distance difference with the cached point cloud). The points), mark the difference areas as new variables (such as new dynamic targets, moving static obstacles).

[0064] Furthermore, incremental risk calculation includes: detecting vehicle position offset via onboard IMU sensors, reinitializing the sensing unit when the offset exceeds a set value; marking the motion state of static obstacles, and performing continuous tracking only on unmarked targets. Specifically, this includes: 1) Vehicle position offset detection. The onboard IMU (MEMS accelerometer and gyroscope) has a sampling rate of 100Hz. The cumulative displacement (obtained through integral acceleration) is calculated from when the vehicle is turned off until it is turned back on. For example, with a set value of 0.5m (threshold for slight vehicle sway), the cumulative displacement along the X / Y axes is... When a significant movement of the vehicle is detected, the sensor units (radar, camera) are triggered to reinitialize (calibrate the zero point and restart parameter configuration).

[0065] 2) Static obstacle marking and tracking.

[0066] First, for reused static obstacles, their displacement is detected by radar for three consecutive frames. Marked as "static" (mark bit = 1); displacement Then it is marked as "dynamic" (mark bit = 0). Subsequently, continuous tracking (updating position and velocity every frame) is performed only on targets with mark bit = 0 (dynamic obstacles) and newly added variables (such as targets not recorded in the cache). Static targets with mark bit = 1 have their positions updated every 5 seconds (to reduce computing power consumption).

[0067] In step S5, when a door closing signal is detected and the risk is eliminated, the active warning system is automatically reset. Specifically, this includes: 1. Set the trigger conditions.

[0068] Door closing signal: The door control switch changes from open to normally closed (lasting >2 seconds, confirming complete closure); Risk cleared: Radar and camera detected no risky targets for 5 consecutive frames (150ms) (static obstacle distance). No dynamic target enters the danger zone; distance from low obstacles. ).

[0069] 2. Reset operation.

[0070] Hardware-wise: Turn off millimeter-wave radar (cut off 12V power supply), camera (stop video acquisition), speaker (mute), LED light strip (turn off when power is off), and haptic feedback unit (stop vibration).

[0071] At the software level: clear local cached environmental data (except for pre-sleep data that needs to be retained in rest scenarios), reset risk level parameters (risk value = 0), and reset the hierarchical decision model (restore the initial state).

[0072] State switching: The system switches from "active" state to "standby" state, retaining only the door opening signal detection function (power consumption reduced to <1W), and waits for the next trigger.

[0073] Example 2 This embodiment discloses a door opening risk proactive warning system based on differentiated hierarchical control.

[0074] like Figure 2 As shown, the door opening risk active warning system based on differentiated hierarchical control includes: The multimodal sensing module includes a millimeter-wave radar array, a wide-angle fisheye camera, and an ultrasonic sensor arranged around the door. The hierarchical decision-making module is used to execute the hierarchical risk assessment model and output risk level codes; The dynamic response module includes a programmable LED ambient lighting strip, a beamforming speaker, and a haptic feedback unit; The scenario adaptation module is used to realize incremental risk calculation and cloud-based collaborative optimization for rest scenarios; The vehicle body controller is used to coordinate the data flow and command transmission of the multimodal sensing module, hierarchical decision-making module, dynamic response module, and scene adaptation module.

[0075] Furthermore, the deployment method of the millimeter-wave radar array includes: installing three radar nodes embedded in the vehicle door, respectively covering monitoring sectors at three different angles: 0-30°, 30-60°, and 60-90°; each radar node uses a TDMA mechanism to transmit frequency-modulated continuous waves in a time-division manner to avoid signal interference.

[0076] like Figure 3 As shown, the dynamic response module also includes an AR-HUD projection unit and a piezoelectric vibrating pad built into the seat; wherein, the AR-HUD projection unit is used to overlay and display obstacle location indicators on the windshield; the piezoelectric vibrating pad built into the seat is used to generate a tactile coding sequence that is positively correlated with the risk level.

[0077] like Figure 4 As shown, the scene adaptation module includes a vehicle micro-motion sensing unit and an environmental change comparison unit; wherein, the vehicle micro-motion sensing unit is used to detect vehicle displacement using a MEMS accelerometer; and the environmental change comparison unit is used to identify scene differences through a feature point matching algorithm.

[0078] Furthermore, it also includes privacy protection mechanisms, namely: embedding a privacy mask layer in real time into the camera video stream to blur faces / license plates in non-risk areas; and automatically overwriting locally stored risk data after encryption within 24 hours.

[0079] Those skilled in the art will understand that the modules or steps of the present invention described above can be implemented using general-purpose computer devices. Optionally, they can be implemented using computer-executable program code, thereby allowing them to be stored in a storage device for execution by a computer device, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. The present invention is not limited to any particular combination of hardware and software.

[0080] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A proactive warning method for door opening risk based on differentiated hierarchical control, characterized in that, include: Real-time monitoring of vehicle status; when the vehicle is detected to be parked and at least one door is triggered, the multimodal sensing unit is activated. The multimodal sensing unit synchronously collects static obstacle height information, dynamic live target behavior trajectory and low obstacle spatial distribution information around the vehicle door, and generates a three-dimensional risk representation. Based on the obtained three-dimensional risk representation, a graded risk assessment model is used to dynamically identify the risk level. The construction of the graded risk assessment model includes: First, establishing a static obstacle height-distance mapping table, i.e., when the obstacle height is higher than the vehicle height and the distance is less than a first threshold, it is judged as high risk; then, constructing a low obstacle feature library, i.e., using a machine learning model to identify the contour features of curbs and guardrails, triggering a scratch warning when the distance is less than a second threshold; finally, deploying a live target behavior prediction model, i.e., calculating the motion vector of a live target in real time based on the YOLOv7 framework and predicting the probability of entering a dangerous area; Differentiated warning commands are generated based on the risk level: for low-risk scenarios, the ambient lights are controlled to output monochromatic steady-state light; for medium- and high-risk scenarios, the ambient lights are driven to execute a multi-frequency flashing sequence and the directional sound field speakers are activated. When a door closing signal is detected and the risk is eliminated, the active warning system is automatically reset.

2. The method for proactive early warning of door opening risk based on differentiated hierarchical control as described in claim 1, characterized in that, The activation of the multimodal sensing unit includes: first, waking up the millimeter-wave radar array corresponding to the target door through the body controller to construct a polar coordinate monitoring network with the door as the origin; and simultaneously activating the wide-angle fisheye camera.

3. The method for proactive early warning of door opening risk based on differentiated hierarchical control as described in claim 1, characterized in that, The execution of the differentiated warning command includes: pre-setting multi-level color gamut encoding in the LED light strip embedded in the car door, with green indicating safety, yellow indicating warning, and red indicating high risk; generating an adjustable focused sound beam through a directional speaker to concentrate sound energy in the direction of risk; and superimposing a pulse vibration warning from a tactile feedback module when a living target is detected approaching rapidly.

4. The method for proactive early warning of door opening risk based on differentiated hierarchical control as described in claim 1, characterized in that, It also includes a rest scenario re-triggering mechanism. Specifically, if the vehicle door experiences a closing and reopening event after the vehicle is turned off, and the seat sensor detects that the occupant is still present, the environmental data collected before hibernation is reused, and incremental risk calculation is only performed on the newly added variables.

5. The method for proactive early warning of door opening risk based on differentiated hierarchical control as described in claim 4, characterized in that, The incremental risk calculation includes: detecting the vehicle position offset through the vehicle-mounted IMU sensor, re-initializing the sensing unit when the offset is greater than a set value; marking the motion state of static obstacles, and performing continuous tracking only on unmarked targets.

6. A door opening risk proactive warning system based on differentiated hierarchical control, employing the door opening risk proactive warning method based on differentiated hierarchical control as described in any one of claims 1-5, characterized in that, include: The multimodal sensing module includes a millimeter-wave radar array, a wide-angle fisheye camera, and an ultrasonic sensor arranged around the door. The hierarchical decision-making module is used to execute the hierarchical risk assessment model and output risk level codes; The dynamic response module includes a programmable LED ambient lighting strip, a beamforming speaker, and a haptic feedback unit; The scenario adaptation module is used to realize incremental risk calculation and cloud-based collaborative optimization for rest scenarios; The vehicle body controller is used to coordinate the data flow and command transmission of the multimodal sensing module, hierarchical decision-making module, dynamic response module, and scene adaptation module.

7. The door opening risk active warning system based on differentiated hierarchical control as described in claim 6, characterized in that, The deployment method of the millimeter-wave radar array includes: installing three radar nodes embedded in the vehicle door, covering monitoring sectors at three different angles: 0-30°, 30-60°, and 60-90° respectively; each radar node uses a TDMA mechanism to transmit frequency-modulated continuous waves in a time-division manner to avoid signal interference.

8. The door opening risk active warning system based on differentiated hierarchical control as described in claim 6, characterized in that, The dynamic response module also includes an AR-HUD projection unit and a piezoelectric vibrating pad built into the seat; wherein, the AR-HUD projection unit is used to overlay and display obstacle location indicators on the windshield; the piezoelectric vibrating pad built into the seat is used to generate a tactile coding sequence that is positively correlated with the risk level.

9. The door opening risk active warning system based on differentiated hierarchical control as described in claim 6, characterized in that, The scene adaptation module includes a vehicle micro-motion sensing unit and an environmental change comparison unit; wherein, the vehicle micro-motion sensing unit is used to detect vehicle displacement using a MEMS accelerometer; and the environmental change comparison unit is used to identify scene differences through a feature point matching algorithm.

Citation Information

Patent Citations

  • Vehicle door anti-collision early warning method and system and related equipment

    CN110103821A

  • Vehicle door opening anti-collision early warning system and method integrating touch sense and vibration

    CN119459515A