A vehicle all-around scratch and chassis protection system and method

Abstract

This invention relates to a comprehensive vehicle anti-scratch and chassis protection system and method. The system includes a distributed 4D millimeter-wave radar and vision sensors, particularly by mounting lateral radar on moving components such as the front wheel steering knuckles, completely eliminating lateral blind spots during steering. Simultaneously, the system deeply integrates cross-modal perception data and innovatively incorporates real-time vehicle dynamics parameters such as vehicle speed and suspension travel to solve for transient suspension compression. This system not only displays obstacle distances with centimeter-level precision but also dynamically predicts inertial collision risks and bottoming-out risks when going down steps, thereby outputting graded warnings or actively implementing deceleration and emergency braking interventions, achieving high-precision three-dimensional active protection for vehicles in all scenarios.

Description

Technical Field

[0001] This invention belongs to the field of vehicle active safety and intelligent driving assistance technology, and relates to a vehicle all-around anti-scratch and chassis protection system and method. Specifically, it relates to a vehicle all-around anti-scratch and chassis protection system and method based on multi-sensor deep fusion. More specifically, this invention relates to a system and method that utilizes 4D millimeter-wave radar (including steering-responsive side radar) and visual sensors for cross-modal environmental collaborative perception, and combines real-time vehicle dynamics models for dynamic inertial risk and elevation clearance prediction, thereby achieving all-weather, high-precision three-dimensional spatial detection and active graded intervention of the vehicle's surroundings and chassis. Background Technology

[0002] With the continuous growth of car ownership, urban road traffic environments are becoming increasingly complex. Vehicles face various risks of scraping and bottoming out in three dimensions when encountering oncoming traffic on narrow roads, passing through width-restricted barriers, entering and exiting underground parking garages, and driving on unpaved roads. Statistics show that over 60% of vehicle scrape accidents occur on the sides of the vehicle and the areas around the front bumper. Underbody collisions (especially damage to the battery pack in electric vehicles) not only incur high repair costs but also pose a significant safety hazard due to battery thermal runaway. Current solutions for protecting the vehicle's overall structure and chassis have the following significant limitations and technical barriers:

[0003] 1. Weak cross-modal perception fusion capability and lack of absolute accuracy in distance display: Traditional ultrasonic radar has a short detection range and low angular resolution, making it difficult to achieve centimeter-level accurate distance measurement; although reversing cameras or panoramic cameras can provide visual assistance, their performance drops sharply in nighttime, strong light, and inclement weather. Existing multi-sensor systems often lack a low-level adaptive fusion mechanism based on environmental disturbance immunity confidence, resulting in the inability to provide drivers with all-weather, highly robust centimeter-level quantized distance display. Drivers still need to rely on visual ambiguity for judgment, which is highly susceptible to misjudgment due to visual illusions.

[0004] 2. Rigid sensor layout leads to "blind spots" in lateral and low-lying areas: Existing side collision avoidance sensors are typically fixedly mounted on both sides of the bumper or below the exterior rearview mirrors, with their detection field of view rigidly bound to the vehicle body. When the vehicle makes a sharp turn, due to the lack of dynamic tracking capabilities, the system cannot detect dynamic obstacles (such as bollards or flower bed edges) on the inside of the curve in advance, easily leading to "inner wheel difference" collisions. At the same time, low-lying obstacles on both sides of the front of the vehicle (such as parking locks or curb stones) are easily overlooked because they are in the fixed vertical field of view blind spot of traditional forward radar.

[0005] 3. Chassis protection relies on static geometric comparison, lacking dynamic prediction capabilities: Existing chassis protection technologies (such as transparent chassis) largely depend on vision or single-point lasers, making it difficult to accurately reconstruct the three-dimensional geometric features of road surface unevenness. More critically, existing systems can only provide "static environmental perception," completely severing the connection with vehicle dynamic parameters. When a vehicle travels rapidly over uneven surfaces, the system cannot combine parameters such as vehicle speed, suspension stiffness, and sprung mass to predict transient suspension depth compression and dynamic ground clearance changes caused by "inertial nose-dive" or "suspension bottoming out," thus completely losing the ability to actively defend against the risk of dynamic inertial scraping.

[0006] 4. Lack of intervention mechanisms for vertical step terrain such as "downhill followed by steps": When a vehicle enters a downhill courtyard or parking lot entrance with steps, the driver cannot accurately judge the dynamic matching relationship between the vehicle's approach angle, longitudinal passing angle, current vehicle pitch angle, and the geometric parameters of the steps by visual inspection alone. Existing automatic emergency braking (AEB) systems are mainly designed for forward obstacles and lack accurate elevation detection, dynamic passability calculation, and dedicated active emergency braking functions for such negative pitch step terrain, making it extremely easy for the front lip to break or the chassis to become stuck.

[0007] 5. Fragmented system architecture and inefficient collaborative decision-making: Existing functions such as front-end anti-scratch, side anti-scratch, chassis transparent view, and hill descent control are usually pieced together from multiple independent electronic control units (ECUs) and scattered sensor subsystems. This distributed architecture not only leads to high hardware costs and complex wiring harnesses, but also makes it impossible to perform deep fusion of multi-source data and cross-domain collaborative decision-making at the feature level, making it difficult to form a unified, efficient, and hierarchical all-round protection strategy.

[0008] Therefore, in view of the technical problems of the above-mentioned existing technologies in the protection of the vehicle body and chassis, such as blind spots in perception, poor anti-interference ability, disconnect between dynamic and static functions and fragmented system architecture, this invention proposes a vehicle all-round anti-scratch and chassis protection system and method based on multi-sensor deep fusion and vehicle dynamics model prediction. Summary of the Invention

[0009] In view of this, in order to solve the problem of scraping and bottoming out caused by blind spots, lack of accurate distance information, and inability to dynamically predict chassis risks when vehicles meet on narrow roads or enter complex terrain (such as downhill steps, road surfaces with potholes or bumps), the present invention provides a vehicle all-round anti-scratching and chassis protection system and method based on multi-sensor deep fusion and vehicle dynamics model prediction.

[0010] To achieve the above objectives, the present invention provides the following technical solution:

[0011] A vehicle all-around anti-scratch and chassis protection system based on multi-sensor fusion, comprising:

[0012] The sensor subsystem includes at least one set of 4D millimeter-wave radar sensors and at least one set of vision sensors;

[0013] 4D millimeter-wave radar sensors are distributed and installed on the inside of the front bumper, the front wheel hub or steering knuckle, and the inside of the rear bumper of the vehicle to acquire high-precision four-dimensional point cloud data including distance, azimuth, pitch, speed and radar cross section (RCS) of the area in front, to the side and behind the vehicle in real time.

[0014] The vision sensors are installed on the upper part of the vehicle's windshield, the exterior rearview mirrors, and the B-pillar to acquire dense visual image information of the vehicle's surrounding environment.

[0015] The signal processing and feature fusion unit is electrically connected to the radar sensor and the vision sensor. It is used to perform spatiotemporal synchronization, coordinate transformation, preprocessing and feature-level deep fusion based on deep neural networks on multi-source sensing data to generate an enhanced environmental perception feature map.

[0016] The distance calculation and enhanced display unit is connected to the signal processing and feature fusion unit. It is used to calculate the precise Euclidean distance between the left and right front corners and the side of the vehicle and the nearest obstacle in real time, and to display the distance information in digital form on the cab display screen with centimeter-level precision.

[0017] The vehicle dynamics parameter acquisition module acquires real-time dynamic parameters, including vehicle speed, steering wheel angle, suspension compression stroke, and vehicle mass, via the vehicle bus.

[0018] The multi-dimensional risk prediction unit is connected to the signal processing and feature fusion unit and the vehicle dynamics parameter acquisition module. Based on the fused perception data and the vehicle dynamics model, it calculates the risk index of front-end collision, side-body collision, dynamic chassis impact, and bottoming out when going down a step.

[0019] The unified decision-making and planning unit is connected to the multi-dimensional risk prediction unit and generates corresponding early warning information or active obstacle avoidance / deceleration path based on the comprehensive risk level and the driver's current operating intention.

[0020] The vehicle control interface is electrically connected to the vehicle's steering, braking, drive, and active suspension actuators. It is used to send control commands to the corresponding actuators to achieve active obstacle avoidance, active speed reduction, and chassis height adjustment.

[0021] As one of the core innovations of this invention, the installation layout of the 4D millimeter-wave radar sensor adopts a heterogeneous scheme with front and rear bumpers and wheel hubs working together:

[0022] 1. The first forward-facing radar is installed in the center of the inner side of the vehicle's front bumper, with its main antenna beam pointing forward. This radar preferably uses a high-end on-chip radar chip cascaded with 48 transmit and 48 receive channels to provide high angular resolution. Its horizontal field of view is no less than 100°, its vertical field of view is no less than 30°, and its maximum detection range is no less than 350 meters, enabling it to accurately identify the three-dimensional contour information of obstacles in front.

[0023] 2. A second set of lateral road detection radars is fixedly installed near the steering knuckles or wheel hub bearing housings on both sides of the front wheels. This radar uses a dual-beam antenna design.

[0024] (1) Road surface detection beam: facing the direction of vehicle movement and at a downward angle of 15°–30° to the horizontal plane, with a vertical beam width of ±15°, used to continuously scan the road surface in the area 5–30 meters in front of the vehicle to obtain three-dimensional point cloud data of road surface unevenness and slope changes.

[0025] (2) Lateral anti-scratch beam: oriented horizontally towards the outer side of the vehicle body, the horizontal beam width covers a range of ±60° on the side of the vehicle body, used to detect the distance to lateral obstacles when meeting oncoming traffic or passing through narrow passages. This lateral detection beam has high-precision measurement capabilities, with a distance measurement accuracy of up to 0.02 meters, an azimuth accuracy of 0.2°, and a pitch accuracy of 0.6°.

[0026] 3. Install a rear radar in the center of the inner side of the rear bumper of the vehicle. Its main beam is tilted downward at a preset angle (10°-20°) to detect obstacles behind and road conditions such as steps and potholes on the reversing trajectory line.

[0027] As a further improvement of the present invention, the lateral radar sensor mounted on the front wheel rotates synchronously with the steering knuckle and the wheel. The system acquires the steering wheel angle signal in real time through the vehicle's CAN bus and dynamically adjusts the perception strategy according to the current steering angle: specifically, when the absolute value of the steering wheel angle exceeds a preset threshold (e.g., 5°), the system automatically increases the update rate of the lateral detection data of the inner steering radar and the risk weight coefficient, thereby realizing the early perception of obstacles on the inside of the curve and the adaptive adjustment of the warning threshold.

[0028] As an alternative installation option, the 4D millimeter-wave radar sensor can be embedded in the gap between the spokes of the vehicle's wheel rim, rotating with the wheel hub. The system provides precise angle-coded signals through wheel speed sensors or additional Hall sensors, combined with the radar's own trigger timing, and utilizes synthetic aperture or scan accumulation technology to achieve a 360° panoramic scan and high-density point cloud reconstruction of the vehicle's surrounding environment with each rotation of the sensor.

[0029] like Figure 1The diagram illustrates the four-layer system architecture of this invention, from top to bottom: a perception layer (4D millimeter-wave radar group + visual sensor group), a data processing layer (spatiotemporal synchronization, feature fusion, distance calculation), a decision-making and planning layer (multi-dimensional risk prediction, unified decision-making), and an interactive execution layer (audio-visual and tactile warning, chassis / braking / steering control). The vehicle all-around scratch and chassis protection method based on the aforementioned vehicle all-around scratch and chassis protection system includes the following steps:

[0030] S1. Acquisition and Preprocessing of Multi-Source Sensing Data

[0031] Multimodal perception data streams of the vehicle's surrounding environment are acquired in real time through forward-facing radar mounted on the front bumper, side radar mounted on the front wheel steering knuckles, rear radar mounted on the rear bumper, and visual sensors installed throughout the vehicle body. The preprocessing includes performing DBSCAN (density-based clustering with noise) clustering filtering on the radar point cloud data to remove outliers, and performing distortion correction and white balance adjustment on the visual images.

[0032] S2, Cross-modal spatiotemporal synchronization and deep feature fusion

[0033] The radar point cloud data obtained in step S1 is aligned with the visual image data at the microsecond level based on GPS timestamps or the system's local clock. Then, using the pre-calibrated intrinsic and extrinsic parameter matrices of each sensor, the radar point cloud is projected onto the pixel coordinate system of the visual image, or the semantic information of the visual image is back-projected onto the three-dimensional space of the radar point cloud, achieving pixel-level to point cloud-level spatial alignment.

[0034] Furthermore, such as Figure 7 As shown, a deep fusion neural network module based on a deformable attention mechanism is constructed. This module takes the BEV features of radar point clouds and the perspective features of visual images as inputs, and generates enhanced lateral obstacle fusion features, front obstacle fusion features, and road surface unevenness fusion features that are robust to occlusion and poor lighting through cross-modal feature interaction and weighted fusion.

[0035] S3, Precise Distance Calculation and Visualization

[0036] Based on the fused features generated in step S2, the nearest valid obstacle target to the vehicle body is extracted. The distance calculation adopts a weighted Kalman filter fusion strategy: the accurate radial distance measurement value provided by 4D millimeter-wave radar, with an accuracy better than 0.02 meters, is used as the main basis for updating the observation. At the same time, the edge detection and disparity estimation results of the visual image are introduced to correct the uncertainty of the lateral boundary of the obstacle, ensuring that the absolute accuracy of the output distance reaches the centimeter level.

[0037] The calculated precise distance information is overlaid as highlighted numbers on the 360° panoramic imaging interface of the dashboard or central control display in the driver's cab. The specific display logic is as follows:

[0038] (1) Safe zone (distance > 50cm): Green numbers are displayed.

[0039] (2) Warning range (30cm ≤ distance ≤ 50cm): Yellow numbers are displayed, and dynamic ripples are superimposed to indicate the warning range.

[0040] (3) Danger zone (distance < 30cm): red numbers flashing.

[0041] Meanwhile, in the 360° panoramic image, colored bands (green / yellow / red) fill the gaps between the vehicle's outline and obstacles, visually displaying the distance relationship between each side of the vehicle and the obstacle. When the vehicle is meeting oncoming traffic on a narrow road or passing through width-restricted barriers, the driver can monitor the precise distance between the left and right sides of the vehicle and the obstacle in real time on the display screen (e.g., "35cm on the left, 28cm on the right").

[0042] S4, Front-end dynamic anti-scratching detection and early warning

[0043] Based on fused data from forward and side radars, particularly the low-profile dot clouds on both sides of the front bumper, the system monitors obstacle distances in real time to the front lip and side fender areas. Combined with steering wheel angle, it predicts the trajectory of the front wheels to determine if the front corner of the vehicle will sweep over an obstacle. When the predicted minimum distance between the front fascia and an obstacle is less than a preset safety threshold, a tiered front fascia anti-scratching warning is triggered for that side.

[0044] 1. When the distance enters the 30-50 cm range, the area will be highlighted in yellow on the display screen and an intermittent alert sound will be emitted;

[0045] 2. When the distance is less than 30 centimeters, the area will be highlighted in red on the display screen and a continuous, rapid alert sound will be emitted;

[0046] 3. If the system detects that the driver has not taken effective avoidance measures, such as failing to straighten the steering wheel or slow down, and the predicted collision time is less than the preset threshold, the system can actively trigger slight braking intervention on one side of the wheels or provide steering resistance torque to assist the driver in correcting the driving trajectory.

[0047] S5, Side-mounted anti-scratching detection and warning system

[0048] Based on a homing steering lateral radar installed at the front wheel steering knuckle, combined with a visual sensor at the B-pillar location to supplement blind spots, it monitors the distance to lateral obstacles in all directions from the A-pillar to the C-pillar in real time. A deep learning-based road boundary segmentation network (a variant of PointNet++) is used to process the 4D millimeter-wave radar point cloud, achieving a point cloud segmentation accuracy of 93% and a median distance error of 0.023 meters. This algorithm can accurately distinguish curbs, walls, guardrails, and oncoming vehicles. The A-pillar is located between the windshield and the front door, the B-pillar between the front and rear doors, and the C-pillar between the rear door and the rear of the vehicle, primarily supporting the roof and forming the safety frame of the passenger compartment.

[0049] When the distance between the side of the vehicle and an obstacle is determined to be less than a preset side safety threshold, a side collision avoidance warning is triggered. The system displays a real-time diagram on the display screen showing the relative position of the vehicle's side profile and the obstacle in a "top-down heat map" format, and precisely marks the nearest point distance with high-contrast numbers.

[0050] S6. Chassis scrape prevention prediction and path planning based on dynamic model

[0051] When an obstacle with a height difference exceeding a preset threshold is detected on the road ahead, the system initiates a dynamic risk assessment:

[0052] 1. Based on road surface point cloud data, a Delaunay triangular mesh optimization algorithm with feature line constraints is used to reconstruct the 3D surface of the road. The height of any bulge or the depth of any depression in the road surface ahead is calculated. ; Obtain information including vehicle speed V and sprung mass Unsprung mass Suspension stiffness Real-time vehicle dynamics parameters, including the shock absorber damping coefficient C, are used, and a pre-set 1 / 4 vehicle dynamics model is introduced. The maximum dynamic compression of the suspension due to inertia when the wheel rolls over an obstacle is calculated by solving the second-order differential equation. :

[0053] (1)

[0054] Calculate the effective ground clearance after considering transient compression. in This refers to the vehicle's static ground clearance; if the conditions are met... If a risk of chassis scraping is detected, a voice prompt will be given saying "The road surface ahead is uneven, please slow down to below X km / h" and the danger zone will be marked on the HUD or instrument panel. If the vehicle is equipped with an intelligent driving assistance system, it can automatically trigger a smooth deceleration intervention.

[0055] 2. Active lane change planning: If there is an obstacle (such as a deep pit more than 15cm deep) that cannot be safely passed in the current lane, and the adjacent lane is confirmed to be safe and free of vehicles by the lateral radar, the system will provide the driver with lane change suggestions in combination with global path planning, or actively plan a smooth lane change and obstacle avoidance trajectory.

[0056] S7, Step-down bottoming detection and emergency braking intervention

[0057] When a step feature with a negative elevation angle is detected on the road ahead, the step height is fitted. and the edge line of the step; read the front overhang length of the vehicle. Approach angle Minimum ground clearance And combined with the actual vehicle pitch angle fed back by the suspension travel sensor Calculate the current ground clearance of the lower edge of the front bumper. Combined with vehicle speed V, the dynamic sag of the front suspension at the moment the front wheels contact the step is estimated. To obtain dynamic ground clearance ;

[0058] The logic for assessing the risk of a safety net follows these rules:

[0059] 1. Failure criteria: If

[0060] (2)

[0061] Therefore, it is determined that even if the static condition passes, there is still a risk of being forced to back up.

[0062] 2. Critical threshold can be determined if: the step height is within the critical range, and the vehicle speed... (e.g., 5km / h) then p is determined to be due to the front suspension being severely compressed by the inertia of the downhill slope, which poses a risk of dynamic bottoming out.

[0063] When it is determined that there is a risk of a bottoming out:

[0064] 3. The system issues a clear voice warning when the distance to the step is 3-5 meters: "The step ahead is too high, there is a serious risk of bottoming out, please stop immediately!"

[0065] 4. If the driver does not take effective braking measures and the longitudinal distance between the vehicle and the step continues to shrink to less than 2 meters, the system will ignore the accelerator pedal signal and automatically trigger a high-priority emergency braking command to bring the vehicle to a reliable stop in front of the step with the maximum safe deceleration.

[0066] 5. If the system determines that the step is passable but the vehicle's attitude needs to be adjusted (e.g., it is recommended to drive at an angle to increase the approach angle), the system will overlay a recommended safe driving trajectory guide line on the display screen to assist the driver in passing safely.

[0067] S8. Unified Decision-Making and Planning with Tiered Proactive Intervention

[0068] In response to the risk assessment result of any one of steps S3 to S7, the system determines the risk index based on the comprehensive risk index. Activate the tiered response mechanism. The comprehensive risk index... It is obtained by weighted summation of each sub-risk index (front, side, chassis, bottom) and its corresponding collision time (TTC).

[0069] Level 1 Response (Alert / Warning): When When in a low-risk zone, the system only displays the location and distance of the risk via the dashboard / HUD and issues a soft audible and visual alarm.

[0070] Level 2 response (auxiliary intervention): when When entering the medium-risk zone and the driver fails to take effective evasive action, the system provides tactile alerts to the driver through the steering wheel vibration motor, seat vibration motor, or seatbelt pretensioner. It can also actively apply a small corrective torque through the electric power steering system to provide the driver with intuitive evasive guidance.

[0071] Level 3 Response (Proactive Intervention): When When the high-risk threshold is reached and the time window is insufficient for the driver to react, the system automatically triggers emergency braking or single-wheel braking intervention to forcibly change the vehicle's trajectory or state in order to actively avoid collisions or scrapes.

[0072] The beneficial effects of this invention are as follows:

[0073] 1. The vehicle all-round anti-scratch and chassis protection system disclosed in this invention makes full use of the high-precision ranging capability of 4D millimeter-wave radar (accuracy better than 0.02 meters), and displays the distance between the vehicle and obstacles in real time and accurately in digital form on the driver's cab display screen, transforming the driver's subjective and vague judgment into objective and accurate data, which greatly reduces the psychological burden and the probability of misjudgment when driving on narrow roads.

[0074] 2. The all-around vehicle scratch prevention and chassis protection system disclosed in this invention achieves a leap from static perception to dynamic, responsive perception through a heterogeneous layout of fixed forward / rear radar and responsive side radar. In particular, the installation scheme of the front wheel steering knuckle completely solves the problem of blind spots in the perception of the inside of curves, realizing active monitoring of scratch and bottoming risks in all scenarios without blind spots.

[0075] 3. The all-around anti-scratch and chassis protection system for vehicles disclosed in this invention not only focuses on the static geometric dimensions of obstacles, but also incorporates dynamic parameters such as vehicle speed, suspension stiffness, and sprung mass into the risk assessment model for the first time. By solving the vibration differential equation in real time, it achieves advanced prediction and active deceleration intervention for dynamic scratch risks such as "inertial nose-diving" and "suspension bottoming out," transforming "passive response" into "active prevention."

[0076] 4. The all-around vehicle scratch protection and chassis protection system disclosed in this invention addresses the common and high-risk scenario of descending slopes and encountering steps. The system utilizes radar's pitch resolution capability to identify the precise height of the step in advance and performs dynamic passability calculations based on the vehicle's attitude. In the event of driver error, the system can decisively execute emergency braking, filling the gap in existing AEB systems' ability to identify and respond to such vertical obstacles.

[0077] 5. The vehicle all-around anti-scratch and chassis protection system disclosed in this invention features a 4D millimeter-wave radar with excellent all-weather operation capabilities. It can effectively penetrate rain, fog, smoke, and dust, and is unaffected by drastic changes in lighting conditions. Its data is deeply fused with visual sensor data at the feature level, forming a highly redundant perception system with complementary advantages, ensuring stable high-performance output even in harsh environments such as nighttime, backlighting, rain, and snow.

[0078] 6. The all-around vehicle scratch protection and chassis protection system disclosed in this invention has mature installation technology and existing wiring harness channels for the sensor installation locations (inner side of the front bumper, front wheel steering knuckle, and inner side of the rear bumper). In particular, the radar at the front wheel steering knuckle can be wired using the existing wheel speed sensor wiring harness channel. The entire solution requires minimal modification, has high engineering feasibility, and is easy to implement for rapid mass production and functional upgrades on existing vehicle platforms.

[0079] Other advantages, objectives, and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination, or may be learned from practice of the invention. The objectives and other advantages of the invention can be realized and obtained through the following description. Attached Figure Description

[0080] To make the objectives, technical solutions, and advantages of the present invention clearer, the preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings, wherein:

[0081] Figure 1 This is a flowchart of the vehicle all-around anti-scratch and chassis protection system of the present invention;

[0082] Figure 2 This is a diagram showing the sensor installation layout and electrical control architecture of the present invention;

[0083] Figure 3 This is a schematic diagram of the installation of the lateral radar of the present invention;

[0084] Figure 4 This is a schematic diagram of the installation of the rearward radar of the present invention;

[0085] Figure 5 This is a scene of two cars meeting on a narrow road.

[0086] Figure 6 A flowchart for handling a downhill slope followed by steps;

[0087] Figure 7 Flowchart for deep fusion of cross-modal features;

[0088] Figure 8 Flowchart for 3D reconstruction of road surface unevenness and chassis risk assessment;

[0089] Figure 9 This is a schematic diagram of adaptive switching of Kalman fusion weights based on dynamic environment confidence.

[0090] Reference numerals: Forward radar 101, left front side radar 102, right front side radar 103, rearward radar 104. Detailed Implementation

[0091] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention.

[0092] Example

[0093] I. Sensor Installation Layout and Electrical Control Architecture

[0094] like Figure 2 As shown, the hardware architecture of the system of the present invention includes: a 4D millimeter-wave radar sensor group (forward radar 101, left front side radar 102, right front side radar 103, and rear radar 104), a vision sensor group (forward binocular camera, left side camera, and right side camera), a domain controller (including a signal processing and feature fusion unit, a distance calculation and display unit, a risk prediction unit, and a unified decision planning unit), as well as actuator interfaces (steering interface, braking interface, drive interface, and suspension interface) and a graded warning module (display module, audio module, and tactile module).

[0095] 1. Front bumper radar installation details:

[0096] The first 4D millimeter-wave forward-facing radar, namely Forward Radar 101, is fixedly installed above the center anti-collision beam on the inner side of the vehicle's front bumper using a special metal bracket. To ensure that the vertical field of view covers the area from the ground to the near field, the radar is installed at a specific height... The radar is preferably positioned 50cm above the ground, with the radar radome normal parallel to the vehicle's longitudinal axis. This radar employs a cascaded design to achieve 48 transmit and 48 receive channels, achieving a horizontal angular resolution of 0.8° and a vertical angular resolution of 1.5°. Its main lobe covers the far-field area directly in front, while a sidelobe gain compensation algorithm ensures high sensitivity for detecting low-lying obstacles in the left and right front corners.

[0097] 2. Front wheel steering knuckle radar installation details:

[0098] like Figure 3 As shown, a compact 4D millimeter-wave radar, namely the left front lateral radar 102 and the right front lateral radar 103, is fixedly mounted on the steering knuckles of the left and right front wheels using the existing brake lines / wheel speed sensor mounting bolt holes and via adapter brackets. This radar employs an AIP (Air-Independent Propulsion) design to reduce size and weight. Internally, a microstrip array antenna design achieves dual-beam functionality, wherein:

[0099] Road surface detection beam: The center of the beam points at a 22° downward angle to the horizontal plane, and the beamwidth (3dB) is ±10°. This beam primarily scans the road surface ahead when the vehicle is traveling in a straight line, and scans the road surface on the inside of the turning side when turning.

[0100] Lateral anti-scratch beam: The center direction is perpendicular to the side of the vehicle, and the beam width (3dB) is ±60°. This beam is responsible for establishing an electronic fence along the side of the vehicle.

[0101] The radar is connected to the vehicle's wiring harness via a flexible coaxial cable and rotates synchronously with the steering knuckle. The system acquires the steering wheel angle signal provided by the EPS in real time via the CAN bus. .when When the system determines that the vehicle is turning, it automatically prioritizes the lateral beam detection data of the inner steering radar (such as the right front lateral radar 103 when turning right) to the highest priority and reduces its risk warning threshold by 15% to compensate for the blind spot changes caused by wheel steering.

[0102] Rear bumper radar installation details:

[0103] like Figure 4 As shown, a rearward 4D millimeter-wave radar, namely rearward radar 104, is fixedly installed in the center of the inner side of the rear bumper of the vehicle. In order to accurately detect ground steps or potholes when reversing, the radar uses a wedge-shaped mounting bracket to tilt its antenna normal downward by 12°, ensuring that the lower edge of the vertical field of view can illuminate the ground 0.5 meters behind the vehicle.

[0104] II. Precise Distance Calculation and Intuitive Visualization Methods

[0105] Taking the front wheel steering knuckle side radar as an example, the point cloud data frame rate output by the left front side radar 102 and the right front side radar 103 is 20Hz. After the signal processing and feature fusion unit performs voxel filtering downsampling on each frame of point cloud, it uses Euclidean clustering algorithm to segment the point cloud into independent targets.

[0106] Step 1: Extraction of multi-source independent observations.

[0107] Select the point closest to the origin of the vehicle coordinate system after clustering to obtain the 4D millimeter-wave radar ranging value. Visual ranging values ​​are extracted within the ROI region corresponding to the visual image using a binocular stereo matching algorithm. Construct the system observation vector at the current time k:

[0108] (1)

[0109] Step 2: Real-time assessment of environmental and target confidence.

[0110] The system calculates a normalized dynamic confidence factor (ranging from 0 to 1, with higher values ​​indicating more reliable data) based on environmental and target characteristics:

[0111] 1. Visual confidence The system extracts the current camera's exposure time parameters, overall image contrast, and wiper operating frequency. When the system detects that the wipers are operating at a high frequency (indicating heavy rain) or that the image contrast has dropped sharply (indicating heavy fog or strong glare), it uses an exponential decay model to reduce the frequency. The value of .

[0112] 2. Radar confidence level The system extracts the variance and Doppler signal-to-noise ratio of the radar cross section (RCS) corresponding to the obstacle. If the RCS value fluctuates very little and the signal-to-noise ratio is high over multiple consecutive frames, then... It approaches 1; if the RCS fluctuates drastically due to multipath effects, it will decrease accordingly. .

[0113] Step 3: Adaptive observation noise covariance matrix update.

[0114] Utilizing radar reference noise variance calibrated in advance through extensive real-vehicle testing With visual reference noise variance Based on the aforementioned dynamic confidence level, a dynamic observation noise covariance matrix is ​​constructed. :

[0115] (2)

[0116] As can be seen from this formula mechanism, when the visual sensor is affected by strong light interference, the confidence level... As the decrease approaches 0, the noise variance term corresponding to the matrix will be amplified nonlinearly and drastically.

[0117] Step 4: Predicting and updating the optimal estimate of state variables.

[0118] Prior estimates of state variables (obstacle distance and relative velocity) calculated based on the vehicle kinematics model. and covariance matrix Then, the Kalman gain with environmental adaptability is calculated. :

[0119] (3)

[0120] Then update the optimal distance state estimate at the current moment:

[0121] (4)

[0122] Figure 9 This diagram illustrates the adaptive switching of Kalman fusion weights based on dynamic environmental confidence. It shows that in the interference range of "strong light / severe weather," the visual confidence weight (solid line) drops precipitously, while the radar confidence weight (dashed line) instantly reaches full compensation. The significant advantage of this algorithm lies in its mathematically implemented immunity to environmental interference. For example, when a vehicle enters a dark underground parking garage and is illuminated by the high beams of an oncoming vehicle, the visual algorithm fails due to overexposure. The sharp decrease causes the corresponding terms in the covariance matrix to tend towards infinity. Kalman gain. The calculation automatically ignores visual observation residuals and instantly transfers 100% of the weights to the 4D millimeter-wave radar's distance prediction. When illumination is restored, the weights smoothly transition back to the multi-sensor equilibrium state through dynamic calculation of the covariance matrix. This mechanism ensures that no matter how abruptly the external environment changes, the distance error output to the distance display and decision-making intervention system is always locked at the absolute centimeter level.

[0123] The distance calculation and display unit uses an adaptive weighted fusion algorithm:

[0124] 1. Radar ranging value Select the point that is closest to the origin of the vehicle coordinate system after clustering, and its distance value is... Confidence level It depends on the signal-to-noise ratio of the point cloud's RCS (radar cross section).

[0125] 2. Visual disparity value Within the region of interest (ROI) corresponding to the visual image, a dense disparity map is calculated using a binocular stereo matching algorithm to extract the visual distance to the edges of corresponding obstacles. Confidence level It depends on the texture richness and lighting conditions of the area.

[0126] 3. Fusion Output: Final Display Distance

[0127] (5)

[0128] At night or in rainy or foggy weather The weighting is reduced, and the weights are automatically tilted towards the radar to ensure stable output.

[0129] like Figure 5 As shown, in a narrow road meeting scenario, the system displays the vehicle's outline on the central control screen in the driver's cab using a 3D free-view or top-down perspective. On the left and right sides of the vehicle outline, distance values ​​(e.g., "32" and "28") are displayed in prominent sans-serif font, close to the vehicle body, with the unit defaulting to centimeters. Simultaneously, a semi-transparent red / yellow / green band is rendered in the ground projection area between the vehicle and the obstacle. When the distance enters the warning zone, the corresponding number changes color from white to yellow and flashes slowly; when it enters the danger zone, the number becomes bright red and flashes rapidly, while the audio module 402 emits a "beep-beep-beep" warning sound with a frequency that increases as the distance decreases.

[0130] III. Bottom-out detection and active emergency braking logic when going down a step

[0131] like Figure 6 As shown, this embodiment describes a detailed processing flow for a downhill step scenario.

[0132] 1. Scene Trigger: When a high-precision map or navigation data indicates that there is a slope ahead. When encountering a downhill section of road with a Point of Interest (POI) such as a parking lot entrance within 20 meters ahead, the system is pre-set to enter "Step-Down Alert Mode". Simultaneously, if the forward-facing radar 101 detects a step change in the z-axis (height) direction of the ground echo ahead, i.e., a step feature, it will also automatically trigger this mode.

[0133] 2. Step Parameter Extraction: The downward beam of the forward-facing radar 101 (or the road surface detection beams of the left and right lateral radars 102 / 103) performs a grid scan of the step area with a vertical accuracy of 1 cm. The upper plane, lower plane, and elevation of the step are fitted using the RANSAC algorithm, and the step height is accurately calculated. And the equation of the step edge line.

[0134] 3. Vehicle passability dynamics calculation:

[0135] The system reads the vehicle's inherent parameters: front overhang length. Approach angle Minimum ground clearance Wheelbase Dynamic ground clearance .

[0136] The system obtains the current actual vehicle pitch angle from the CAN bus. (Calculated by the suspension travel sensor) and the current vehicle speed V.

[0137] Calculate the current ground clearance of the lower edge of the front bumper:

[0138] (6)

[0139] Calculate the limiting approach angle condition required to pass through the steps:

[0140] (7)

[0141] in This is the horizontal distance from the front wheel contact point to the edge of the step.

[0142] By combining the vehicle speed V, the dynamic sag of the front suspension at the moment the front wheels contact the step can be predicted. The amount of subsidence can be obtained by looking up a table, which is obtained from vehicle dynamics simulation or actual vehicle calibration, and records the maximum suspension compression at different vehicle speeds and slopes.

[0143] 4. Risk Decision-Making and Execution:

[0144] Condition A (Serious Risk): If

[0145] (8)

[0146] It was determined that it was not safe to pass, among which This refers to the dynamic ground clearance.

[0147] Warning: When the vehicle is 5 meters away from the step, an audio message will be played via the audio module: "Warning! The step ahead is too high and there is a serious risk of bottoming out. Please stop immediately!" A red warning box will pop up on the central control screen.

[0148] Intervention: If the distance is reduced to 2 meters and the driver does not press the brake pedal, the unified decision planning unit immediately sends an emergency braking request of -8m / s² to the braking interface and cuts off the drive torque output, forcing the vehicle to stop in front of the edge of the step.

[0149] Condition B (can be used with caution): If

[0150] (9)

[0151] That is, the steps are relatively high but theoretically passable, among which This represents the dynamic deflection of the front suspension at the moment the front wheel contacts the step.

[0152] Guidance: The system calculates a recommended path to approach the step at an angle (this path effectively increases the vehicle's approach angle). On the 360° panoramic view, the system draws a yellow guide trajectory line, guiding the driver to approach the step at an angle of approximately 30°-45°. Simultaneously, a text prompt reads, "Please proceed slowly at an angle along the guide line."

[0153] IV. Coordination of Dynamic Inertial Scratching Prediction and Active Deceleration

[0154] When the system detects a speed bump or protrusion with a height of Δh ahead through the road surface three-dimensional reconstruction module (step S6 of claim), the inertial collision prediction process is initiated. Figure 8 This is a flowchart illustrating the 3D reconstruction of road surface unevenness and chassis risk assessment. It specifically demonstrates how, after radar point clouds are reconstructed using Delaunay triangulation, the elevation difference is input into the "Dynamic Model Risk Assessment" module, which includes dynamic parameters (M, K, C, V), ultimately determining whether to reduce speed or change lanes.

[0155] 1. Predictive Calculation Based on Vehicle Dynamics Model: The system calls a preset vehicle dynamics model (in this implementation case, the vehicle dynamics model adopts a 1 / 4 vehicle vibration transfer function model). Input the current vehicle speed V and sprung mass. Unsprung mass Suspension stiffness The damper damping coefficient C is also considered. The maximum dynamic suspension compression generated by the vehicle body is estimated by solving the second-order differential equation of the vehicle dynamics model. It is expressed as a multivariable function:

[0156] (10)

[0157] 2. Risk Comparison: Calculate the current effective ground clearance

[0158] (11)

[0159] in This refers to the vehicle's static ground clearance. If the obstacle height... If a risk of chassis scraping is detected, the system will immediately trigger an automatic speed reduction to a safe speed threshold. Down.

[0160] 3. Intelligent speed reduction suggestions and intervention:

[0161] (1) Solve the vibration model in reverse order of the system to obtain the maximum safe speed that can be allowed to pass through the protrusion. .

[0162] (2) The system displays a prompt on the instrument panel or HUD: "Road surface bump ahead, please slow down." "Below km / h".

[0163] (3) If the vehicle is in ACC (Adaptive Cruise Control) or AP (Automatic Driving Assist) mode, the system can automatically and smoothly reduce the set speed to [the desired speed]. It resumes its original speed after passing obstacles, and the entire process requires no driver intervention, achieving comfortable active protection.

[0164] Alternatives and Variations

[0165] It is understood that there are multiple alternative implementations of the technical solution of this invention that do not depart from its core idea:

[0166] 1. Radar Selection and Performance Alternatives: The 4D millimeter-wave radar is not limited to a specific brand or number of channels. Any millimeter-wave radar capable of providing the target's three-dimensional spatial coordinates (x, y, z) and velocity information, such as an imaging radar employing DDS technology, can be applied to the system architecture of this invention.

[0167] 2. Adaptable Fine-tuning of Installation Position: In addition to the steering knuckle, the front wheel side radar can also be installed in fixed positions on the lower control arm, shock absorber strut, or wheel arch liner, and its position can be compensated for by software algorithms caused by steering. The rear radar can also be configured with two corner radars installed on either side of the rear bumper to obtain a wider rearward and side field of view.

[0168] 3. Diverse presentation of human-machine interface: In addition to the central control screen, distance information can be flexibly integrated into the 3D vehicle model of HUD head-up display, streaming media rearview mirror or full LCD instrument panel.

[0169] 4. Personalized customization of warning thresholds: All parameters involved in this invention, such as distance thresholds, volume levels, and tactile feedback intensity, can be calibrated and adjusted through the vehicle settings menu, and can be linked to the user's personal account to achieve a personalized safety protection experience for each individual.

[0170] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A vehicle all-around anti-scratch and chassis protection system, characterized in that, include: The sensor subsystem includes multiple sets of 4D millimeter-wave radar sensors for acquiring point cloud data of the inside of the curve and the side of the vehicle in real time, and multiple sets of visual sensors for acquiring dense visual image information of the environment around the vehicle. The signal processing and feature fusion unit is electrically connected to the sensor subsystem and is used to perform spatiotemporal synchronization and feature fusion on multi-source sensing data. The vehicle dynamics parameter acquisition module is used to acquire real-time vehicle dynamics parameters, including vehicle speed, steering wheel angle, and suspension compression stroke. The multidimensional risk prediction unit, connected to the signal processing and feature fusion unit and the vehicle dynamics parameter acquisition module, is configured to: calculate the dynamic compression of the vehicle suspension in real time based on the fused three-dimensional road surface perception data, combined with the vehicle dynamics model and the current vehicle speed and suspension stiffness parameters, and predict the dynamic chassis collision risk and the risk of bottoming out when going down a step, combined with the vehicle's static ground clearance. The unified decision-making and planning unit and the vehicle control interface generate control commands based on the comprehensive risk level predicted by the multi-dimensional risk prediction unit and send them to the vehicle actuators to implement graded early warning or proactive intervention.

2. The vehicle all-around anti-scratch and chassis protection system as described in claim 1, characterized in that, The sensor subsystem includes at least a lateral radar installed in the steering knuckle or wheel bearing housing area of ​​the front wheel of the vehicle. The lateral radar deflects synchronously with the steering of the wheel. The vision sensor is installed on the upper part of the windshield, the exterior rearview mirror and the B-pillar.

3. The vehicle all-around anti-scratch and chassis protection system as described in claim 2, characterized in that, The side radar installed in the front wheel steering knuckle or wheel hub bearing area adopts a dual-beam antenna design, including: a road surface detection beam, facing the direction of vehicle travel and with a downward tilt angle, used to scan the uneven geometry of the road surface in front of the vehicle and the inner steering area; and a side anti-scratch beam, facing the outer side of the vehicle horizontally, used to construct a distance detection electronic fence on the side of the vehicle.

4. The vehicle all-around anti-scratch and chassis protection system as described in claim 3, characterized in that, When performing distance calculation, the signal processing and feature fusion unit uses an adaptive Kalman filter algorithm based on dynamic environment confidence.

5. The vehicle all-around anti-scratch and chassis protection system as described in claim 4, characterized in that, When predicting the risk of bottoming out on a step, the multi-dimensional risk prediction unit performs the following operations: obtains the height of the step in front through radar sensors; obtains the vehicle's front overhang length, minimum ground clearance, and approach angle, and calculates the dynamic front ground clearance by combining the real-time vehicle pitch angle and the dynamic sag estimated by the vehicle speed; when it is determined that the bottoming out risk conditions are met, it triggers active emergency braking or plans an oblique passage trajectory.

6. A method for all-around vehicle scratch prevention and chassis protection based on the all-around vehicle scratch prevention and chassis protection system described in claims 2-5, characterized in that, Includes the following steps: S1. Environmental data is acquired through distributed 4D millimeter-wave radar sensors and vision sensors, especially blind spot data obtained through lateral radars mounted on the front steering knuckles of the vehicle. The points closest to the origin of the vehicle's coordinate system after clustering are selected to obtain the 4D millimeter-wave radar ranging values. Visual ranging values ​​are extracted within the ROI region corresponding to the visual image using a binocular stereo matching algorithm. Construct the system observation vector at the current time k: ; S2. Perform time alignment and spatial feature fusion on cross-modal sensing data; evaluate the environmental immunity confidence of the visual sensor in real time. and the signal confidence of millimeter-wave radar ; The observation noise covariance matrix is ​​dynamically updated based on the confidence level. : in, and The baseline noise variances for radar and vision are respectively used; the adaptive Kalman gain is calculated to update the optimal distance state estimate of the obstacle. S3. Based on the dynamic environment confidence level, calculate the precise distance between the vehicle body and obstacles, and display it visually on the display terminal. The final displayed distance is: At night or in rainy or foggy weather, The weighting is reduced, and the weights are automatically tilted towards the radar to ensure stable output; S4. Based on the fusion data of forward radar and side radar, predict whether the minimum distance between a certain side of the front face and an obstacle is less than the preset safety threshold, so as to trigger the corresponding side of the front face anti-scratching grade warning. S5: Based on the follow-up steering side radar installed at the front wheel steering knuckle, combined with the visual sensor at the B-pillar position to supplement the visual blind spot, it monitors the distance of all-round side obstacles from the A-pillar to the C-pillar in real time, judges whether the distance between the side of the vehicle and the obstacle is less than the preset side safety threshold, and triggers the side anti-scratch warning. S6. Reconstruct the three-dimensional curved surface of the road ahead, input the obstacle height difference into the vehicle dynamics model, and solve the transient dynamic suspension compression. If the sum of the obstacle height and the dynamic suspension compression is greater than the static ground clearance, the chassis protection active deceleration is triggered. S7. Detect the terrain where the downhill slope meets the steps, fit the geometric model of the steps and combine it with the dynamic attitude of the vehicle body to calculate the passability, and perform active braking or trajectory guidance. S8. Based on the multi-dimensional risk assessment results, output corresponding display, audible and visual warnings, or braking / steering control commands.

7. The method for all-around anti-scratch and chassis protection of vehicles as described in claim 6, characterized in that, Step S6 uses the road surface point cloud data from forward and side radars to reconstruct the three-dimensional surface of the road surface using the Delaunay triangular mesh optimization algorithm with feature line constraints.

8. The method for all-around anti-scratch and chassis protection of a vehicle as described in claim 7, characterized in that, Step S6 calls the preset vehicle dynamics model, specifically the multi-dimensional risk prediction unit 1 / 4 vehicle vibration transfer function model, and inputs the current vehicle speed V and sprung mass. Unsprung mass Suspension stiffness And the damper damping coefficient C, by solving the second-order differential equation, the maximum dynamic suspension compression generated by the vehicle body is estimated. ; Calculate the effective ground clearance considering transient compression ;in This refers to the vehicle's static ground clearance; if the obstacle height meets the conditions... If the system detects a risk of inertial collision, it will issue a voice prompt saying "The road surface ahead is uneven, please slow down to below X km / h" and mark the danger zone on the HUD or instrument panel. If the vehicle is equipped with an intelligent driving assistance system, it will automatically trigger a smooth deceleration intervention.

9. The method for all-around anti-scratch and chassis protection of a vehicle as described in claim 8, characterized in that, Step S7, the calculation of vehicle dynamic attitude passability, specifically involves: when a step feature with a negative pitch angle is detected on the road surface ahead, the step height is fitted. And the edge line of the step, read the vehicle's inherent parameters: front overhang length Approach angle Minimum ground clearance Wheelbase ; and combined with the actual vehicle pitch angle fed back by the suspension travel sensors. Accurately calculate the dynamic ground clearance of the lower edge of the vehicle's front bumper. ; Calculate the limiting approach angle condition required to pass through the steps: , in The horizontal distance from the front wheel contact point to the edge of the step; combined with the vehicle speed V, the dynamic sag of the front suspension at the moment the front wheel contacts the step is estimated. To obtain dynamic ground clearance ; Based on spatial triangular geometric projection relationships, the bottom-line risk assessment logic is executed: if the failure condition is met: If the system determines that there is a risk of bottoming out, it will issue a voice warning at a preset safe distance. If the longitudinal distance is less than the set braking threshold, it will automatically trigger a high-priority longitudinal emergency braking command to force the vehicle to stop.

Images