Anti-dazzle light control method and system based on multi-sensor confidence and trajectory prediction

Abstract

This invention relates to the field of vehicle headlight anti-glare control technology, and particularly to an anti-glare supplementary lighting control method and system based on multi-sensor confidence and trajectory prediction. The control method includes the following steps: S1, collecting environmental information from multiple sensors and unifying spatial coordinates; S2, constructing a FoV confidence grid, counting the number of overlapping coverages of multiple sensors in each grid, calculating the regional fusion confidence, and quantifying the perception reliability of different spaces; S3, fusing multi-sensor targets to obtain the current state quantity; S4, predicting the target and the vehicle's future trajectory, and calculating the anti-glare risk within the future time window; S5, determining the supplementary lighting priority area based on the vehicle's future trajectory and the reliable environmental area, and generating the supplementary lighting area in conjunction with anti-glare constraints; S6, jointly deciding the brightness of each light area, smoothing the result, and outputting it to the matrix, pixel, and DLP modules. Using the control method and system of this invention improves the stability, continuity, and safety of lighting in complex road environments.

Description

Technical Field

[0001] This invention relates to the field of vehicle headlight anti-glare control technology, and in particular to an anti-glare supplementary lighting control method and system based on multi-sensor confidence and trajectory prediction. Background Technology

[0002] With the development of technologies such as intelligent headlights, ADB (Adaptive High Beam), pixel headlights, and DLP projection headlights, vehicle lighting systems have gradually evolved from simple high / low beam switching to "perception-driven lighting." Modern vehicle lighting systems need to dynamically adjust the lighting area and shading pattern based on environmental information such as objects ahead, road curvature, lane markings, oncoming vehicles, vehicles in front, and pedestrians, in order to ensure the driver's visibility while avoiding glare. Current mainstream anti-glare headlight control solutions include the following categories: (1) Anti-glare control based on a single forward-facing camera: The camera identifies the taillights of the vehicle in front, the lights of oncoming vehicles, and lane lines to determine whether to turn off the high beams or create a glare-proof zone. This solution is quite sensitive to weather, ambient light, road surface reflection, and occlusion, and its stability is insufficient at night and in complex scenarios.

[0003] (2) Cornering illumination based on steering signals and vehicle speed: The driving direction is calculated using steering wheel angle, yaw rate and vehicle speed, and the headlights are turned or the inner lighting of the curve is extended. This scheme only reflects the movement trend of the vehicle itself and cannot fully perceive external dynamic targets, and has limited fitting of the actual geometric shape of the road.

[0004] (3) Headlight control based on ADS perception output: The target, lane line, and curvature information output by the intelligent driving domain controller are directly transmitted to the headlight controller to generate high beam control, anti-glare and shading, and cornering lighting commands. Although this scheme enhances perception capabilities, it usually uses perception results directly and lacks joint modeling of differences in sensor coverage capabilities, differences in the credibility of FoV overlap areas, and the future movement trend of the target.

[0005] (4) Target detection-triggered ADB zone control: Based on the target's azimuth, distance, and category, determine the light zones that need to be turned off or have their brightness reduced to achieve local anti-glare. However, most systems are based on the target's current position and belong to "passive response control", which has a response lag for targets that are rapidly approaching, cutting into, or entering / exiting curves.

[0006] The aforementioned existing technologies have the following main shortcomings: a. Insufficient robustness in complex environments: The recognition accuracy of a single or limited sensing source decreases significantly under conditions of rain, fog, backlight, or obstruction, which can easily lead to unstable anti-glare control. b. Area perception capability is not quantified and utilized: the sensor output is assumed to be reliable, and the differences in perception reliability at FoV edges, overlapping areas and different spatial areas are not considered, resulting in an overly aggressive or conservative control strategy. c. Insufficient foresight: Anti-glare and supplemental lighting based on the target position at the current moment have lags in shading or supplemental lighting in scenarios such as high-speed oncoming traffic, vehicles cutting in front, and curve entrances and exits; d. Lack of a unified coordination mechanism between anti-glare and supplementary lighting: Anti-glare focuses on "turning off where" while supplementary lighting focuses on "illuminating where", lacking a unified decision-making framework, which can easily lead to fragmented light patterns and discontinuous lighting; e. Lack of predictive light pattern generation mechanism: Adjustments are usually triggered only after the target has entered the illuminated or risky area, failing to fully utilize multi-sensor fusion and trajectory prediction capabilities to achieve early control.

[0007] Therefore, there is an urgent need to propose a predictive anti-glare supplementary lighting control method that integrates multi-sensor information, spatial area confidence, and target trajectory prediction to solve the problems of insufficient robustness of single sensors, inconsistent regional control accuracy, response lag, and the disconnect between anti-glare and supplementary lighting, thereby improving the stability, continuity, and safety of lighting in complex road environments. Summary of the Invention

[0008] The technical problem to be solved by this invention is: in order to solve the problems of insufficient robustness of single sensors, inconsistent regional control accuracy, response lag, and disconnect between anti-glare and supplementary lighting in the existing technology, this invention provides an anti-glare and supplementary lighting control method and system based on multi-sensor confidence and trajectory prediction.

[0009] The technical solution adopted by this invention to solve its technical problem is: a multi-sensor confidence and trajectory prediction anti-glare supplementary lighting control method, comprising the following steps: S1. Multiple sensors collect environmental information and unify spatial coordinates; S2. Construct a FoV confidence grid, count the number of multi-sensor overlap coverages in each grid, calculate the regional fusion confidence, and quantify the perception reliability in different spaces. S3. Fuse targets from multiple sensors to obtain the current state variables; S4. Predict the target and the future trajectory of the vehicle, and calculate the anti-glare risk within the future time window; S5. Determine the priority area for supplementary lighting based on the future trajectory of the vehicle and the reliable area of ​​the environment, and generate the supplementary lighting area in combination with anti-glare constraints; S6, jointly decides the brightness of each lamp zone, smooths it, and outputs it to the matrix, pixel, and DLP module.

[0010] S1 introduces multi-source sensing to reduce the risk of failure of a single sensor in scenarios such as rain, fog, and low light. S2 quantifies the sensing reliability of different spatial regions to provide a basis for regional differentiation in subsequent control. S4 shifts from responding to the current moment to predicting the future moment to identify risks in advance. S5 and S6 handle anti-glare and supplementary lighting in a unified framework to avoid light pattern fragmentation. S4, S5, and S6 form a prediction-evaluation-adjustment closed loop to achieve advance light pattern adjustment.

[0011] Furthermore, spatial coordinate unification is performed in S1, specifically as follows: First, a spatial transformation is performed on the local coordinate systems of different sensors to unify the position measurement benchmark. The position measurement vector in the sensor coordinate system is represented as follows: Its position in the vehicle coordinate system is represented as , in, and These are the components of the measurement position vector in the sensor coordinate system. and These are the components of the target position vector in the vehicle coordinate system. Let be the installation yaw angle of the i-th sensor relative to the vehicle coordinate system. and Let i be the installation position of the i-th sensor relative to the vehicle reference point; Next, rotate the velocity vector to unify the reference direction of motion. The velocity vector is then represented as: Its representation in the vehicle coordinate system is as , in, and The velocity component of the target relative to the sensor. and The velocity component of the target relative to the sensor in the vehicle coordinate system; Finally, all sensor measurements of environmental targets are mapped to the same vehicle coordinate system to form a unified target observation set.

[0012] Multi-sensor data must be in the same coordinate system to be effectively fused; otherwise, multiple sensors are just multiple isolated information sources. Unifying the spatial coordinates ensures that multi-source data can be fused. Only after unifying the coordinate system can the FoV overlap coverage be counted between different sensors, providing a prerequisite for spatial alignment for confidence grid construction.

[0013] Furthermore, the FoV confidence grid is constructed in S2, specifically including: First, define the modeling area in front of and around the vehicle as follows: , Where x and y are the positions of a point in the grid coordinate system. and Boundary values ​​for mesh modeling; Set the grid vertical resolution to The horizontal resolution of the grid is Calculate any point Corresponding two-dimensional grid coordinates The calculation formula is: , Where floor() is the floor function; Then use the one-dimensional index conversion formula Linearize the two-dimensional grid into a unique index, where, Vertical grid index , Horizontal grid index , Total number of vertical grids ; Finally, a grid structure with fast table lookup was constructed to record the sensor coverage, overlap, regional confidence level, and historical control status of each spatial unit.

[0014] Discretizing a continuous space into grid cells makes the differences in sensing capabilities between different regions recordable, queryable, and computable. The grid structure records historical control states, providing state memory for subsequent smooth control and predictive decision-making.

[0015] Furthermore, S2 calculates the number of multi-sensor overlap coverages for each grid, specifically including: The spatial location of each grid center is calculated from the FoV grid index using the following formula: , ; Then, the coordinates are transformed to the coordinate systems of each sensor. The calculation formula is as follows: , in, and The coordinates of the target grid center in the sensor coordinate system. It is the transpose of the rotation matrix. and These represent the sensor's longitudinal and lateral positions, respectively. Calculate the polar coordinates of the grid center in the sensor coordinate system: the distance r of the grid center relative to the sensor and the azimuth deflection angle θ. in, , , If r and θ both fall within the sensor FoV boundary, then the center of the grid is determined to be covered by the sensor. Finally, the number of radar, camera, and modal coverages of each grid is counted to generate a FoV overlap map.

[0016] By statistically analyzing how many and how many types of sensors cover each grid, a FoV overlap map is generated to identify reliable sensing areas and areas with blind spots or weak coverage. Multi-sensor overlap areas typically have higher reliability, and identifying these areas in this step can enhance control capabilities in complex scenarios.

[0017] Furthermore, based on the FoV overlap plot, a FoV overlap confidence score is introduced, specifically: First, define the region confidence function related to distance and azimuth: Confidence function related to azimuth angle , Distance-related confidence function ; in, Here, is the Sigmoid function, a represents the steepness of the angle decay adjustment, b represents the offset of the Sigmoid function in the angular direction, and c represents the distance decay coefficient. Further define the regional confidence function ; Then calculate the overall FoV overlap confidence for the same grid. The expression is: , in, The total number of radar sensors. This represents the total number of camera sensors. and These are the confidence functions for the k-th radar and the m-th camera, respectively. and It is an adjustable parameter for Sigmoid angle decay. It is an adjustable parameter for exponential distance decay; and It is an adjustable parameter for Sigmoid angle decay. It is an adjustable parameter for exponential distance decay; Further, a normalized form is introduced, and the expression is: , Finally, the fusion confidence score is obtained by combining sensor health status, target consistency, and scene stability. : , in, , , and To calibrate the weights, It is the confidence level of the sensor's health status. It is the confidence level of scene stability.

[0018] In addition to counting the number of coverage points, it also combines factors such as distance, azimuth angle, sensor health status, and scene stability to output a comprehensive and quantifiable fusion confidence score, directly solving the reliability issues of existing systems such as default sensor output; the introduction of sensor health status and scene stability further improves robustness in complex environments.

[0019] Furthermore, S3 specifically includes: After completing spatial alignment and confidence calculation, the observations of the targets from each sensor are fused. The fused data yields at least the following state variables for each target: , Where x represents the longitudinal position of the target in the vehicle coordinate system, and y represents the lateral position of the target in the vehicle coordinate system. The target's longitudinal relative velocity, The target's lateral relative velocity; At the same time, the corresponding fusion covariance matrix P and fusion confidence are retained for this objective. .

[0020] The fused target state integrates information from multiple sensors, making it more accurate and stable than that of a single sensor. The fusion confidence level is used in subsequent anti-glare risk assessments, and the anti-glare risk of targets with low confidence levels will be weighted accordingly to avoid false triggering.

[0021] Furthermore, in S4, the predicted target and the vehicle's future trajectory are as follows: For the future A constant velocity model is used for first-order prediction, and the predicted target state is: , , in, For the goal of the future moment The predicted vertical position, For the goal of the future moment The predicted lateral position, Let be the vertical position of the target at time t. Let be the horizontal position of the target at time t. For the prediction time interval; The future lighting center trajectory of the vehicle is characterized by combining the vehicle's curvature prediction, and the expression is: , in, This represents the longitudinal path independent variable in the road coordinate system / vehicle coordinate system. This represents the lateral offset at the vertical position x. , , and These are the fitting parameters obtained based on lane lines, road curvature, and vehicle motion state. Finally, the future fill light center line and the boundary of the high-brightness area of ​​the vehicle headlights are determined along the predicted trajectory.

[0022] Furthermore, considering the effects of target acceleration or road curvature, in future moments... The second-order prediction model is used to predict the target state as follows: , , in, For the longitudinal acceleration component of the target, The acceleration component of the target in the lateral direction.

[0023] By shifting the target position from the current moment to the target position in the future moment, the system can know in advance which lighting zone the target will enter and when it will enter; the vehicle trajectory prediction provides a directional basis for the generation of the supplementary lighting area, so that the supplementary lighting is no longer blindly shining forward, but a forward-looking illumination along the driving trajectory.

[0024] Furthermore, S4 calculates the anti-glare risk within future time windows, specifically including: Define the goal at a future moment Anti-glare risk index for: , in, To integrate confidence levels for the target future moments. The degree of overlap between the target future location and the high-brightness area of ​​the light. Weights for target type; Define future time windows Comprehensive anti-glare risk within for: , when Greater than the preset threshold In such cases, the corresponding light area is directly shaded or its brightness reduced to achieve predictive anti-glare.

[0025] The risk at a single point in time is extended to the cumulative risk over a future time window, avoiding misjudgments caused by brief occlusion or detection jitter, while enabling early response to rapidly approaching targets; the output of anti-glare risk is directly input into the joint decision of S6, and coordinated with the supplementary lighting requirements.

[0026] Furthermore, the fill light priority area in S5 for: , in: The area representing the vehicle's future driving path; Areas where there are no high-level anti-glare risk targets within the future time window; This represents a high-confidence region where the FoV overlap confidence level is greater than a threshold. In curve scenarios, priority should be given to forward illumination of the inner side of the curve, the road exit direction, and the area near the edge of the lane; in oncoming traffic scenarios, priority should be given to maintaining high beam illumination on the non-glare side of the target.

[0027] Anti-glare constraints are explicitly defined as a prerequisite for supplementary lighting generation, resolving the contradiction between illumination and anti-glare within the same framework; the supplementary lighting area includes a high-confidence region with FoV overlap confidence greater than the threshold, ensuring that supplementary lighting only occurs in areas with reliable perception, avoiding ineffective supplementary lighting in blind spots or low-confidence areas; the supplementary lighting area is based on the vehicle's future driving path, which is forward-looking, rather than just based on the current vehicle orientation.

[0028] Furthermore, S6 specifically refers to: First, the brightness of each light zone is determined jointly based on the predicted anti-glare zone and the supplementary lighting zone. For the i-th light zone, its output brightness is defined. for: , in, The base brightness of the i-th light zone, This represents the anti-glare suppression amount corresponding to the i-th light zone. This represents the supplementary lighting enhancement amount corresponding to the i-th lighting zone. and These are the weighting coefficients; Apply a smoothing constraint to the brightness of the lamp area at adjacent time intervals: , in, Let η be the decision brightness of the i-th light zone at time t, and η be the smoothing factor. For the time before the i-th light zone 'output brightness' For the current decision-making brightness, Finally, the headlight controller outputs the brightness of each light zone to the matrix LED module, pixel module, or DLP projection module to achieve predictive anti-glare and supplemental lighting control.

[0029] Anti-glare (suppression) and supplemental lighting (enhancement) work together in the same brightness calculation formula to achieve a truly unified decision, rather than a simple splicing of two independent modules; smoothing constraints ensure that predictive adjustment will not cause light pattern flickering or abrupt changes due to drastic changes between frames, improving lighting comfort; the smoothing factor η enables the system to achieve a balance between fast response and stable transition.

[0030] A system for implementing the anti-glare supplementary lighting control method based on multi-sensor confidence and trajectory prediction as described above includes: The environmental perception module collects information about the environment around the vehicle through a forward-facing camera and a forward-facing millimeter-wave radar. The data preprocessing and spatial alignment module performs spatial unification of sensor coordinates, mapping the target observation data of each sensor to the same vehicle coordinate system to form a unified target observation set; The FoV mesh and overlap confidence calculation module constructs a FoV lookup table-style mesh, counts the number of overlaps of each mesh covered by multiple sensors, and calculates the region fusion confidence based on distance, azimuth angle, sensor health status, and scene stability. The target fusion and trajectory prediction module fuses targets from multiple sensors to obtain target state variables and predicts the future state of the targets and the vehicle's driving trajectory. The anti-glare risk assessment module calculates the anti-glare risk index of the target at future moments, accumulates the comprehensive anti-glare risk within the future time window, and determines whether the target has entered a bright or dazzling area. The supplementary lighting area generation module determines the priority area for supplementary lighting based on the future trajectory of the vehicle and the reliable area of ​​the environment, and generates the supplementary lighting area under the condition of meeting the anti-glare constraint. The headlight control module jointly decides the brightness of each light zone based on the anti-glare zone and the supplementary light zone, and outputs the result to the matrix, pixel or DLP module after smoothing constraint to achieve predictive anti-glare and supplementary light control.

[0031] The beneficial effects of this invention are: This invention effectively solves the problems of decreased recognition accuracy, target loss, and unstable control caused by a single sensor in complex scenarios such as nighttime, rain, fog, backlight, and occlusion by introducing multi-sensor fusion and spatial coordinate unification, and significantly improves the accuracy of target recognition and environmental perception. This invention solves the problem of inconsistent perception reliability in existing technologies, such as the default sensor output and unused areas, by constructing a FoV confidence grid, statistically counting the number of overlapping coverages of multiple sensors, and combining distance, azimuth angle, sensor health status, and scene stability to calculate the fusion confidence. This makes anti-glare and supplementary lighting control more accurate and reasonable in different spatial areas. This invention, by introducing target and vehicle future trajectory prediction, defines the comprehensive anti-glare risk within a future time window, solving the problems of glare lag, supplementary lighting delay, and insufficient response to rapidly approaching targets caused by the target position at the current moment in existing solutions, and realizing the early identification of glare risks and pre-adjustment of light patterns; This invention solves the problems of anti-glare constraint and supplementary lighting area generation in the existing solution by placing anti-glare constraint and supplementary lighting area generation in the same decision chain and using joint brightness decision and smoothing constraint. This solves the problems of anti-glare control and supplementary lighting control being separated, light pattern discontinuity, and local light spot jump in the existing solution. While avoiding glare to other traffic participants, it improves the lighting effect and driving comfort in key areas. The control method and system proposed in this invention are applicable to various intelligent headlight platforms such as ADB headlights, matrix headlights, pixel headlights, and DLP projection headlights, and have good platform compatibility, engineering application prospects, and industrialization promotion value. Attached Figure Description

[0032] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0033] Figure 1 This is a structural block diagram of the system of the present invention.

[0034] Figure 2 This is a control block diagram for the external communication of the system of this invention.

[0035] Figure 3 This is a control architecture diagram of a specific example of the system of the present invention.

[0036] Figure 4 This is a flowchart of the method of the present invention. Detailed Implementation

[0037] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0038] Example 1: like Figure 1 As shown, this embodiment provides an anti-glare supplementary lighting control system based on multi-sensor confidence and trajectory prediction, including: The environmental perception module collects information about the environment around the vehicle through a forward-facing camera and a forward-facing millimeter-wave radar. The data preprocessing and spatial alignment module performs spatial unification of sensor coordinates, mapping the target observation data of each sensor to the same vehicle coordinate system to form a unified target observation set; The FoV mesh and overlap confidence calculation module constructs a FoV lookup table-style mesh, counts the number of overlaps of each mesh covered by multiple sensors, and calculates the region fusion confidence based on distance, azimuth angle, sensor health status, and scene stability. The target fusion and trajectory prediction module fuses targets from multiple sensors to obtain target state variables and predicts the future state of the targets and the vehicle's driving trajectory. The anti-glare risk assessment module calculates the anti-glare risk index of the target at future moments, accumulates the comprehensive anti-glare risk within the future time window, and determines whether the target has entered a bright or dazzling area. The supplementary lighting area generation module determines the priority area for supplementary lighting based on the future trajectory of the vehicle and the reliable area of ​​the environment, and generates the supplementary lighting area under the condition of meeting the anti-glare constraint. The headlight control module jointly decides the brightness of each light zone based on the anti-glare zone and the supplementary light zone, and outputs the result to the matrix, pixel or DLP module after smoothing constraint to achieve predictive anti-glare and supplementary light control.

[0039] The environmental perception module overcomes the limitations of a single sensor by leveraging the complementary characteristics of multiple sensor types, providing rich and redundant raw environmental data for subsequent processing. The data preprocessing and spatial alignment module first filters, denoises, and unifies the format of the raw perception data to improve data quality. Signal processing and filtering algorithms are used to remove measurement noise and outliers, ensuring the validity and consistency of the input data and reducing processing errors in subsequent modules. Then, data from different sensors are unified into the same vehicle coordinate system to achieve spatial consistency. A coordinate transformation matrix is ​​used to convert the measurements from the local coordinate systems of each sensor to a unified vehicle reference system, resolving the data heterogeneity issue caused by different sensor installation positions and orientations, and providing a spatial reference for subsequent mesh construction and target fusion. The FoV mesh and overlap confidence calculation module first constructs a FoV mesh model of the road area to describe the field of view coverage of each sensor. It discretizes the continuous space in front of and around the vehicle into grid cells with fixed resolution and establishes a mapping relationship between grid index and spatial position. This transforms the continuous spatial perception problem into a computable discrete grid problem, providing a structured carrier for regional confidence quantification. Then, it calculates the perception confidence of different sensors in the overlapping area to improve the reliability of target recognition. By statistically analyzing how many radars, cameras, and multimodal sensors jointly cover each grid cell, and combining distance attenuation, angle attenuation, sensor health status, and scene stability, it outputs a comprehensive fusion confidence score to quantify the perception reliability of different spatial areas and solve the problem of default sensor reliability in existing technologies. The target fusion and trajectory prediction module first fuses target information from multiple sensors to generate a unified target state result. Using weighted fusion or Kalman filtering algorithms, it integrates observations of the same target from multiple sensors, outputting the optimal estimated position and velocity, improving the stability and accuracy of target detection and providing a reliable target state for trajectory prediction. Then, it predicts future trajectories based on the target's historical motion state, providing a basis for advance control. It employs constant velocity models, constant acceleration models, or kinematic models considering road curvature to extrapolate the target's position at future times, giving the system forward-looking capabilities and solving the lag problem of traditional solutions based on current-time control. The anti-glare risk assessment module analyzes whether the target may be illuminated by vehicle headlights, assesses the glare risk level, calculates the spatial overlap between the predicted target position and the high-brightness area of ​​the headlights, and combines target type weights and fusion confidence to accumulate a comprehensive risk index within the future time window. This triggers anti-glare actions before the target enters the glare area, achieving predictive anti-glare. The supplementary lighting area generation module generates areas that enhance illumination while avoiding glare. It performs logical intersection calculations on the vehicle's future driving path area, the target area without anti-glare risk, and the high FoV confidence area, and performs geometric offset optimization for special scenarios such as curves and oncoming traffic to ensure that the supplementary lighting is both effective and safe, and avoids glare to others.Then, considering both anti-glare and supplementary lighting requirements, a final headlight control strategy is generated. This strategy simultaneously considers anti-glare suppression and supplementary lighting enhancement within the same brightness decision equation, coordinating their priorities through weighting coefficients. This addresses the issues of disconnected anti-glare and supplementary lighting, and discontinuous light patterns, present in existing solutions. The headlight execution control module converts the control strategy into headlight drive commands, achieving dynamic light pattern control. A smoothing filter is applied to perform first-order hysteresis filtering on the brightness decisions of adjacent frames, suppressing light pattern jumps and ensuring visual comfort during light pattern changes, avoiding flickering or abrupt changes. Finally, based on the control results, the corresponding light pattern is output, achieving precise illumination and anti-glare supplementary lighting. By independently controlling the switching and brightness of each LED matrix unit or DLP micromirror array, pixel-level light pattern adjustment is achieved, converting control commands into actual light fields to complete the final lighting execution.

[0040] like Figure 2 As shown, this embodiment illustrates the communication topology and data interaction flow between the control system and other electronic control units of the vehicle. It defines the message format and transmission cycle between each node based on the vehicle network protocol, clarifies the external data dependencies of the system and the transmission path of control commands, and ensures the collaborative work between the controllers.

[0041] The data interaction interface between the control system and other controllers in this embodiment includes: IVI (In-vehicle Infotainment System) is used to send projection animations and interactive commands to the vehicle lighting controller. It transmits DLP projection content and human-machine interaction instructions through the vehicle Ethernet or USB channel to realize advanced interactive functions such as personalized light shows and welcome projections. The BCM (Body Controller) is used to send wake-up commands to the left and right front light controllers and forward intelligent driving information. As the gateway node of the vehicle network, the BCM routes signals such as vehicle speed, gear position, light switch, and steering wheel angle on the CAN bus to the headlight controller, providing the system with necessary vehicle status information and enabling remote wake-up in low-power mode.

[0042] ADS (Intelligent Driving Controller) is used to identify and process information such as external environmental targets, lane lines, and road edges, and transmits the coordinate information to the vehicle lighting controller in the form of structured data. Through deep learning target detection algorithms and sensor fusion algorithms, it outputs information such as traffic participants, lane lines, and drivable areas.

[0043] L_LCU / R_LCU (Left / Right Front Light Control Driver) is installed with the vehicle lights and is used to control and drive the left and right front lights. It receives brightness control commands from the real-time computing unit and controls the output of the LED matrix or DLP module through PWM modulation or current driving. As the hardware carrier of the vehicle light execution control module, it converts the digital commands of the decision layer into physical light output.

[0044] The left and right headlights consist of headlight-related functional light panels, DLP modules, and vehicle light structural components. The LED or laser light source on the light panel is reflected by optical lenses or DLP micromirrors to form a specific light distribution pattern. As the final light output actuator, it converts electrical signals into lighting patterns that meet regulatory requirements.

[0045] like Figure 3 As shown, taking the left light controller (L_LCU) as an example, its internal hardware modules and their interconnections adopt an embedded system architecture, integrating communication interfaces, real-time computing units, storage units, and various driver units into the same hardware platform, providing a hardware operating environment that meets functional safety level (ASIL B) and real-time requirements. The internal hardware of the vehicle light controller includes: The communication unit mainly includes CAN (FD) communication and Ethernet communication, providing the controller with external communication capabilities, realizing data interaction and signal control between sensors, domain controllers and the entire workshop, realizing physical layer signal conversion through CAN transceiver and Ethernet PHY chip, and working with MAC layer controller to complete the sending and receiving of data packets; The real-time computing unit, as the core control center, uses an ASIL B functional safety level MCU chip to handle communication with the vehicle, vehicle lighting communication control, vehicle lighting algorithm, DLP control algorithm, and fault diagnosis. The storage unit, including ROM and RAM, is used to store internal data and code, and the address mapping for code execution and data reading and writing is implemented through the memory controller; The motor drive unit is used to drive the motor inside the headlight. It is controlled by the real-time computing unit and uses an H-bridge circuit to achieve closed-loop control of the position and speed of the stepper motor or brushed DC motor, thereby realizing the mechanical adjustment functions of the headlight, such as horizontal adjustment (curving steering), vertical adjustment (automatic height leveling), and motion control of the ADB sunshade.

[0046] The fan drive unit drives the fan and is controlled by the real-time computing unit. It controls the speed of the brushless DC fan through PWM speed regulation and forms a closed-loop temperature control strategy with temperature sensor feedback to ensure the heat dissipation management of high-power LEDs or DLP modules, ensure that optical components work stably within a safe temperature range, and extend the system life.

[0047] Figure 2 for Figure 1 Provide external data input, Figure 3 for Figure 1 Provide hardware operating environment, Figure 1 guide Figure 2 Interface design Figure 3 The allocation of computing power.

[0048] Example 2: like Figure 4 As shown, an anti-glare supplementary lighting control method based on multi-sensor confidence and trajectory prediction includes the following steps: Step 1: During vehicle operation, multiple sensors collect information about the surrounding environment of the vehicle. The sensors include at least a forward-facing camera and a forward-facing millimeter-wave radar, and may further include side cameras, surround-view cameras, corner radar, lidar, or ultrasonic radar. The collected environmental information includes at least the longitudinal position of the target, the lateral position of the target, the longitudinal relative velocity of the target, the lateral relative velocity of the target, the target category, lane line information, road boundary information, road curvature information, wall surface, obstacle, and projectable area information.

[0049] Since the data output by different sensors resides in their respective local sensor coordinate systems, and the sensors are installed in different positions and orientations, it is necessary to first unify their spatial coordinates. Specifically: First, a spatial transformation is performed on the local coordinate systems of different sensors to unify the position measurement benchmark. The position measurement vector in the sensor coordinate system is represented as follows: Its position in the vehicle coordinate system is represented as , in, and These are the components of the measurement position vector in the sensor coordinate system. and These are the components of the target position vector in the vehicle coordinate system. Let be the installation yaw angle of the i-th sensor relative to the vehicle coordinate system. and Let i be the installation position of the i-th sensor relative to the vehicle reference point; Next, rotate the velocity vector to unify the reference direction of motion. The velocity vector is then represented as: Its representation in the vehicle coordinate system is as , in, and The velocity component of the target relative to the sensor. and The velocity component of the target relative to the sensor in the vehicle coordinate system; Finally, after completing the coordinate transformation, all sensor measurement data of environmental targets are mapped to the same vehicle coordinate system, forming a unified target observation set.

[0050] Specific example: When a pedestrian is approximately 30m in front of the vehicle and 2m to its left, the forward-facing camera can detect the target's position and category information in its own coordinate system. For example, the detected target position is approximately 28m longitudinally and -1.5m laterally. Simultaneously, the forward-facing millimeter-wave radar can detect the target's distance and relative velocity information in its own coordinate system. For example, the detected target position is approximately 31m longitudinally and -2.2m laterally, with a relative velocity close to 0. Because the forward-facing camera and the forward-facing millimeter-wave radar have different installation positions and yaw angles, their output data reside in different local sensor coordinate systems. Therefore, it is necessary to first perform rotation and translation transformations on the position vectors measured by each sensor relative to the vehicle coordinate system based on their installation yaw angles and position parameters, and then perform coordinate rotation transformations on the velocity vectors. This allows the target measurement results output by the forward-facing camera and the forward-facing millimeter-wave radar to be uniformly mapped to the same vehicle coordinate system, forming a unified target observation set. After coordinate unification, the target's fused position in the vehicle coordinate system can be determined to be approximately 30m ahead and 2m to the left, with a relative speed close to 0. The system then maps the target's location to the corresponding field-of-view grid cell in front of the vehicle and analyzes the sensor field-of-view overlap and regional confidence level of that grid cell. Since the target is located within the shared field-of-view overlap area of ​​the forward-facing camera and the forward-facing millimeter-wave radar, and its distance is moderate with a small azimuth deviation, the fusion confidence level of the corresponding region is high. If the target is a pedestrian, the system can determine that it belongs to a traffic participant with high priority for anti-glare control and accordingly implement shading or brightness reduction control on the pixel light area or ADB zone corresponding to the target's direction, while maintaining necessary supplementary lighting in non-sensitive areas on both sides of the target to balance anti-glare safety and road lighting effects. If the target is a vehicle ahead, the system can also dynamically adjust the range, position, and duration of the shading area based on its relative speed and subsequent trajectory prediction results to improve the accuracy and smoothness of anti-glare supplementary lighting control.

[0051] Step 2: To characterize the differences in perception capabilities across different spatial regions, a discretized FoV mesh needs to be constructed in front of and around the vehicle. This includes: First, define the modeling area in front of and around the vehicle as follows: , Where x and y are the positions of a point in the grid coordinate system. and Boundary values ​​for mesh modeling; Set the grid vertical resolution to The horizontal resolution of the grid is Calculate any point Corresponding two-dimensional grid coordinates The calculation formula is: , Where floor() is the floor function; To facilitate table lookup and quick location, a one-dimensional index conversion formula is then used. Linearize the two-dimensional grid into a unique index, where, Vertical grid index , Horizontal grid index , Total number of vertical grids ; Finally, a grid structure with fast table lookup was constructed to record the sensor coverage, overlap, regional confidence level, and historical control status of each spatial unit.

[0052] When performing FoV overlap statistics, it is necessary to obtain the actual spatial location of each grid center and count the multi-sensor overlap coverage of each grid, specifically including: The spatial location of each grid center is calculated from the FoV grid index using the following formula: , ; For any grid center Further determination is needed to determine whether it falls within the field of view of a certain sensor. This requires switching from the vehicle coordinate system to the sensor coordinate system. The calculation formula is as follows: , Its expansion is: , in, and The coordinates of the target grid center in the sensor coordinate system. It is the transpose of the rotation matrix. and These represent the sensor's longitudinal and lateral positions, respectively. The rotation angle; Next, calculate the polar coordinates of the grid center in the sensor coordinate system: the distance r of the grid center relative to the sensor and the azimuth deflection angle θ. in, , , If r and θ both fall within the sensor FoV boundary, then the center of the grid is determined to be covered by the sensor. Finally, the number of radar, camera, and modal coverages of each grid is counted to generate a FoV overlap map.

[0053] Furthermore, based on the FoV overlap map, a FoV overlap confidence score is introduced to quantify the perceived reliability of different regions. Specifically: First, define the region confidence function related to distance and azimuth: Confidence function related to azimuth angle , Distance-related confidence function ; in, Here, is the Sigmoid function, a represents the steepness of the angle decay adjustment, b represents the offset of the Sigmoid function in the angular direction, and c represents the distance decay coefficient. Further define the region confidence function: ; This function reflects the following pattern: the farther the target is from the sensor and the greater its deviation from the sensor's main line of sight, the lower the perception reliability of that area.

[0054] Then calculate the overall FoV overlap confidence for the same grid. The expression is: , in, The total number of radar sensors. This represents the total number of camera sensors. and These are the confidence functions for the k-th radar and the m-th camera, respectively. and It is an adjustable parameter for Sigmoid angle decay. It is an adjustable parameter for exponential distance decay; and It is an adjustable parameter for Sigmoid angle decay. It is an adjustable parameter for exponential distance decay; To adapt to real-world systems, a normalized form is further introduced, and the expression is: , Finally, the fusion confidence score is obtained by combining sensor health status, target consistency, and scene stability. : , in, , , and To calibrate the weights, It is the confidence level of the sensor's health status. It refers to the scene stability confidence level. In this way, the system no longer controls the headlights solely based on "whether the target is detected," but can identify "which credible area the target appears in, whether the current confidence level is high enough, and whether it is worthwhile to block or supplement the light in advance."

[0055] Specific Example: In one embodiment, the vehicle's surrounding environment can be divided into discretized FoV mesh regions. The modeling range in front of and around the vehicle can be set to satisfy -150m ≤ x ≤ 150m longitudinally and -15m ≤ y ≤ 15m laterally, with the mesh longitudinal resolution dx set to 1m and the mesh lateral resolution dy set to 1m. Taking a region 40m in front of the vehicle and 3m away from the vehicle's centerline as an example, the corresponding two-dimensional mesh coordinates can be obtained according to the mesh coordinate calculation formula. Subsequently, a unique mesh index is generated through mesh linearization, and a corresponding one-dimensional FoV mesh lookup table structure is established. Further, according to the mesh center inverse calculation formula, the center position of the corresponding mesh cell is approximately (40.5m, 3.5m). Then, the system transforms the mesh center from the vehicle coordinate system to the coordinate systems of each sensor and calculates its distance r and azimuth angle θ relative to each sensor. When the distance r and azimuth angle θ simultaneously satisfy the corresponding sensor's FoV boundary constraints, it is determined that the mesh center is covered by the corresponding sensor. For example, for forward-facing cameras and forward-facing millimeter-wave radars installed at the front of the vehicle, since the aforementioned area is located within the near-to-medium distance range in front of the vehicle and has a small lateral offset, this area usually falls within the field of view coverage of both the forward-facing camera and the forward-facing millimeter-wave radar, thus forming a multi-sensor FoV overlap area. The system further calculates the comprehensive FoV overlap confidence of this grid area based on the distance attenuation function, the azimuth attenuation function, and the number of FoV overlaps of each sensor. Because this area is at a moderate distance and has a small deviation from the sensor's main viewing axis, its area confidence is high; conversely, for areas that are farther away or close to the sensor's FoV edge, their area confidence is correspondingly lower. Finally, when a target is located within the aforementioned high-confidence FoV overlap area, the system can determine that the target is a reliable sensing target and prioritize triggering anti-glare shading or supplemental lighting control in the corresponding area; while for targets located in low-confidence areas, its control priority is reduced or the control triggering time is delayed, thereby reducing the risk of false shading or false supplemental lighting caused by misidentification and improving the accuracy, stability, and environmental adaptability of ADB headlight control.

[0056] Step 3: After completing spatial alignment and confidence calculation, fuse the observations of the targets from each sensor. After fusion, at least the following state variables for each target should be obtained: , Where x represents the longitudinal position of the target in the vehicle coordinate system, and y represents the lateral position of the target in the vehicle coordinate system. The target's longitudinal relative velocity, The target's lateral relative velocity; At the same time, the corresponding fusion covariance matrix P and fusion confidence are retained for this objective. .

[0057] Specific example: Assume the fusion system simultaneously tracks two targets: Target A is a vehicle traveling in the same direction ahead of the vehicle, located approximately 40m in front of the vehicle and offset laterally by approximately 0.8m; Target B is a traffic participant to the left front, located approximately 65m in front of the vehicle and offset laterally by approximately -3.5m. In one embodiment, the forward-facing millimeter-wave radar first acquires the distance and relative speed information of Target A and Target B, and the forward-facing camera acquires the position contour information and category information of Target A and Target B. Subsequently, according to the aforementioned unified spatial coordinate method, the measurement results from each sensor coordinate system are transformed into a unified vehicle coordinate system, and combined with the overlap coverage and regional confidence of the FoV grid areas where each target is located, the observation results from different sensors are correlated, clustered, state predicted, and fused. After fusion, the fused state variables for target A are: longitudinal position approximately 39.8m, lateral position approximately 0.9m, longitudinal relative velocity approximately -4m / s, and lateral relative velocity approximately 0.2m / s. Simultaneously, the fused state variables for target B are: longitudinal position approximately 64.5m, lateral position approximately -3.3m, longitudinal relative velocity approximately 18m / s, and lateral relative velocity approximately -0.6m / s. Furthermore, the system retains the corresponding fusion covariance matrix P and fusion confidence for targets A and B respectively. Target A, being located within the high-confidence area jointly covered by the forward-facing camera and the forward-facing millimeter-wave radar, and exhibiting good observation continuity and high data consistency, has a smaller corresponding fusion covariance matrix and a higher fusion confidence. Target B, due to its relatively greater distance and larger lateral offset, although still within the effective coverage area of ​​some sensors, has a relatively larger corresponding fusion covariance matrix and a lower fusion confidence than target A. Based on the fusion results, the system can further perform trajectory prediction for the two targets separately: For target A, it can be predicted that it will remain within the vehicle's forward lighting area for a short period of time, thus implementing continuous anti-glare shading or local brightness reduction control in its area; For target B, if its future trajectory is predicted to enter the vehicle's high-brightness illumination area within a preset time window, the system can pre-shade the pixel light area or ADB partition in the direction of target B in advance, thereby avoiding passive response after the target actually enters the main lighting area. Therefore, the system can not only output the position and velocity states of each target in the vehicle coordinate system, but also simultaneously output the covariance and fusion confidence corresponding to these state variables, using them as important bases for subsequent anti-glare decisions, supplementary lighting decisions, and control priority allocation, thereby improving the foresight, accuracy, and stability of vehicle lighting control.

[0058] Step 4: To achieve predictive anti-glare and supplemental lighting, the control method in this embodiment does not only use the target's current state, but further predicts its future trajectory. Specifically: For the future A constant velocity model is used for first-order prediction, and the predicted target state is: , , in, For the goal of the future moment The predicted vertical position, For the goal of the future moment The predicted lateral position, Let be the vertical position of the target at time t. Let be the horizontal position of the target at time t. For the prediction time interval; The future lighting center trajectory of the vehicle is characterized by combining the vehicle's curvature prediction, and the expression is: , in, This represents the longitudinal path independent variable in the road coordinate system / vehicle coordinate system. This represents the lateral offset at the vertical position x. , , and These are the fitting parameters obtained based on lane lines, road curvature, and vehicle motion state. Finally, the future fill light center line and the boundary of the high-brightness area of ​​the vehicle headlights are determined along the predicted trajectory.

[0059] Furthermore, considering the effects of target acceleration or road curvature, in future moments... The second-order prediction model is used to predict the target state as follows: , , in, For the longitudinal acceleration component of the target, The acceleration component of the target in the lateral direction.

[0060] For each predicted future target, determine whether it will enter the high-brightness area or potential glare area of ​​the headlights within a future time window, and calculate the anti-glare risk at future moments, specifically including: Define the goal at a future moment Anti-glare risk index for: , in, To integrate confidence levels for the target future moments. The degree of overlap between the target future location and the high-brightness area of ​​the light. Weights for target type; Furthermore, define the future time window. Comprehensive anti-glare risk within for: , when Greater than the preset threshold At this time, the system does not wait for the target to actually enter the current glare zone, but instead performs shading or dimming treatment on the corresponding light area in advance to achieve predictive anti-glare.

[0061] Specific example (curved road meeting scenario): In one embodiment, when the vehicle is traveling along a curved road at night, the system identifies lane lines, road boundaries, and road curvature information using a forward-facing camera, and continuously detects and tracks oncoming vehicles using forward-facing millimeter-wave radar and corner radar. After completing spatial alignment, FoV overlap confidence calculation, and multi-sensor fusion, the system can obtain the oncoming vehicle's current position, lateral offset, longitudinal relative speed, lateral relative speed, fusion covariance matrix P, and fusion confidence in the vehicle coordinate system. Further, based on the oncoming vehicle's current state, the system uses a constant velocity model or a second-order prediction model to predict the target trajectory within a preset future time window. Simultaneously, combining lane lines, road curvature, and vehicle motion state, the system fits the future lighting center trajectory of the vehicle to determine the future supplementary lighting center line and the boundary of the high-brightness area. When predictions indicate that an oncoming vehicle will enter the vehicle's headlight high-brightness area or potential glare area along a curve trajectory within a future time window, and the target has a high fusion confidence level and a large overlap with the future high-brightness area, the overall anti-glare risk calculated by the system exceeds a preset threshold. In this case, the system does not wait for the oncoming vehicle to actually enter the current glare area, but instead proactively implements shading or brightness reduction control on the corresponding pixel light area or ADB zone it will enter in the future. Simultaneously, the system can still maintain necessary supplementary lighting along the predicted curve trajectory of the vehicle to the inside of the curve, the curve exit, and the road edge area, thus balancing anti-glare effect with visibility of the curved road surface. In other words, in this scenario, the system does not passively prevent glare based solely on the target's current location, but rather uses forward-looking control combined with the future trajectory and future lighting distribution, thereby improving the response lead and control smoothness in curve-crossing scenarios.

[0062] Specific example (rapid entry scenario): In another embodiment, when a vehicle in an adjacent lane suddenly and rapidly enters the vehicle's lane, the forward-facing and side-facing cameras can detect its lateral displacement trend, while millimeter-wave radar or corner radar can detect its distance and relative speed changes. After unifying the spatial coordinates of the data output by each sensor, the system, combined with the FoV grid overlap coverage and regional confidence, performs correlation, fusion, and trajectory management on the target, obtaining the target's fusion state quantity and corresponding fusion confidence in the vehicle coordinate system. Further, the system predicts the target's position at several future moments based on its current longitudinal position, lateral position, longitudinal relative speed, and lateral relative speed. When the prediction indicates that the target will laterally enter the high-brightness illumination area of ​​the vehicle's headlights from an adjacent lane within a short period, and the target's future position has a high degree of overlap with the boundary of the high-brightness area, the anti-glare risk index for the corresponding future moment increases. Within the future time window, if the overall anti-glare risk exceeds a set threshold, the system pre-shields or pre-reduces the brightness of the illumination area that the target will occupy in the future, rather than waiting until the target actually enters the high-brightness area before performing the light zone switching. Furthermore, if the target entry process is accompanied by detection jitter, short-term trajectory oscillation, or temporary mismatch of some sensors, the system can still combine the fusion covariance matrix P, fusion confidence, and scene stability to smoothly constrain the control output, thereby avoiding frequent flashing or repeated on / off switching of the headlights. Thus, in rapid entry scenarios, the system can make advance decisions based on the target's future trajectory and future anti-glare risks, achieving preventative suppression of potential glare risks and improving the continuity, accuracy, and comfort of headlight control.

[0063] Step 5: After meeting the anti-glare constraints, the control method in this embodiment further generates a supplementary lighting area based on the vehicle's future trajectory and the reliable area of ​​the environment, with a priority supplementary lighting area. for: , in, The area representing the vehicle's future driving path; Areas where there are no high-level anti-glare risk targets within the future time window; This represents a high-confidence region where the FoV overlap confidence level is greater than a threshold. In curve scenarios, priority is given to providing forward illumination to the inner side of the curve, the road exit direction, and the area near the lane edge; in oncoming traffic scenarios, priority is given to maintaining high beam illumination on the non-glare side of the target, in order to balance safety and lighting continuity.

[0064] Specific example (nighttime curve scenario) In one embodiment, when the vehicle is traveling at night along a winding mountain road, a forward-facing camera identifies road boundaries, lane lines, and road curvature information, while millimeter-wave radar detects the distance and relative speed of a target vehicle ahead. The system further establishes a target fusion trajectory and the vehicle's future trajectory based on the multi-sensor fusion results. Combining the road curvature, steering wheel angle, yaw rate, and vehicle speed information within a future time window, the system predicts the vehicle's future lighting centerline and generates a future high-brightness area for the high beam.

[0065] At the same time, the system predicts the future trajectory of oncoming vehicles and calculates the degree of overlap between the target's future position and the future bright area. When the prediction results indicate that the oncoming vehicle will enter the potential glare area within the future time window, the corresponding area is marked as a high-level anti-glare risk area, and local shading control is executed in advance.

[0066] After completing the anti-glare constraint, the system is further based on: The future trajectory area of ​​the vehicle; A safe zone where no high-risk targets exist; The FoV overlap has a high confidence level in the region of confidence. Generate a priority area for supplemental lighting.

[0067] In this scenario, since the inside of the curve and the direction of the curve exit are key areas for the vehicle's future driving, the system prioritizes maintaining the high beam illumination intensity in these areas, while continuously supplementing the lighting on the road edges, near the guardrails, and in potential pedestrian areas; and for areas predicted to be occupied by oncoming vehicles, the brightness of the corresponding ADB pixel light area is reduced in advance.

[0068] Therefore, in nighttime curve scenarios, this embodiment can not only suppress potential glare risks in advance, but also maintain continuous lighting in key areas of the curve along the future trajectory of the vehicle, thereby improving the driver's visibility of curve exits, obstacles and road edges.

[0069] Specific example (meeting oncoming traffic) In another embodiment, when the vehicle is traveling on a two-way road at night, the system detects the light area of ​​oncoming vehicles and the lane position through a forward-facing camera, obtains the target distance, longitudinal relative speed and lateral offset through millimeter-wave radar, and completes multi-sensor fusion by combining the results of the corner radar and the visual target association.

[0070] The system further predicts the target's future trajectory based on its current motion state and, combined with the vehicle's future lighting distribution, calculates the potential anti-glare risk within a future time window. When the prediction indicates that an oncoming vehicle will enter the vehicle's high-beam brightness area, the system proactively applies partial glare shielding to the corresponding ADB area.

[0071] At the same time, the system does not completely shut off the entire high beam area, but further adjusts it based on:

[0072] Generate supplementary lighting area.

[0073] In this oncoming traffic scenario, the system prioritizes maintaining high beam illumination in the following areas: Non-glare side area for oncoming vehicles; This vehicle's lane edge area; There are no risk areas further ahead on the road. Areas where pedestrians, bicycles, and obstacles may appear on the right side of the road.

[0074] For example, when an oncoming vehicle is on the left side of the road, the system only dynamically shades the corresponding ADB light area on the left, while continuing to maintain high beam illumination on the right lane edge and the far end of the road. If the oncoming vehicle deviates slightly along the center line of the road, the system combines the fusion covariance matrix and FoV overlap confidence to smoothly adjust the shading boundary to avoid frequent flickering of the light area.

[0075] Therefore, in nighttime oncoming traffic scenarios, this embodiment can not only meet the anti-glare requirements, but also maintain the high beam coverage of non-risk areas to the maximum extent, thereby improving the driver's environmental perception and lighting continuity during oncoming traffic.

[0076] Step Six: Perform joint control based on the predictive anti-glare zone and the supplementary lighting zone to generate the target brightness value for each lighting zone. First, determine the brightness of each lighting zone based on the joint control of the predictive anti-glare zone and the supplementary lighting zone. For the i-th lighting zone, define its output brightness. for: , in, The base brightness of the i-th light zone, This represents the anti-glare suppression amount corresponding to the i-th light zone. This represents the supplementary lighting enhancement amount corresponding to the i-th lighting zone. and This is a weighting coefficient; it increases when a target is expected to enter the corresponding dazzling light zone in the future. When a certain area belongs to a future driving critical area with high credibility and no high-risk targets, increase... .

[0077] To avoid abrupt changes in light pattern, a smooth constraint is applied to the brightness of the lamp area at adjacent time points: , in, Let η be the decision brightness of the i-th light zone at time t, and η be the smoothing factor. For the time before the i-th light zone 'output brightness' For the current decision-making brightness, Finally, the headlight controller outputs the brightness of each light zone to the matrix LED module, pixel module, or DLP projection module to achieve predictive anti-glare and supplemental lighting control.

[0078] Specific examples (conflict scenarios) In one embodiment, when the vehicle is traveling on a two-way curved road at night, the system uses a forward-facing camera to identify the curvature of the curve, the edge of the lane and the area of ​​oncoming vehicle lights, uses millimeter-wave radar to obtain the distance to oncoming vehicles, longitudinal relative speed and lateral offset, and generates the target's future trajectory and fusion confidence through multi-sensor fusion.

[0079] At the same time, the system generates the future high beam brightness area and the future fill light priority area based on the vehicle's steering wheel angle, yaw rate and future path prediction results.

[0080] In this scenario, the system predicts that: (1) Oncoming vehicles will enter the dazzling area corresponding to the left front ADB light zone within a future time window; (2) The exit of the curve and the right edge of the road are the key driving areas for the vehicle in the future; (3) The FoV overlap confidence level in the above areas is high, and no high-risk targets were detected.

[0081] Therefore, the system performs joint control on different light zones: (1) Increase the anti-glare suppression amount for the left-side light area that is predicted to be occupied by oncoming vehicles. Reduce the brightness of the corresponding light area; (2) Increase the supplementary lighting intensity for the lighting areas corresponding to the exit of the curve and the right edge of the road. Maintain high beam brightness; (3) For the intermediate transition area, the brightness change range is dynamically adjusted according to the future risk level and the fusion covariance matrix.

[0082] For example, in the third to fifth light zones on the left, because the predicted future trajectory of the target highly overlaps with the dazzling area, therefore the corresponding... As the light intensity increases, the brightness of the light zone gradually decreases; while light zones 8 to 12 on the right correspond to the future road exit directions, therefore... Increase the beam size to maintain continuous high-beam illumination.

[0083] Therefore, in this conflict scenario, the system does not simply shut down the entire high beam area, but makes a joint decision based on future anti-glare risks and future supplemental lighting needs, thereby simultaneously taking into account anti-glare safety and long-distance visibility on curves.

[0084] Specific example (smoothing effect) In another embodiment, when an oncoming vehicle continues to approach along a curved road section, due to slight lateral swaying of the target detection, changes in road curvature, and vibrations in some sensor detections, if the ADB light zone is controlled directly based on the instantaneous detection results, it is easy to cause local light zones to frequently turn on or off, thereby causing the driver to perceive obvious flashing or light pattern changes.

[0085] Therefore, this embodiment further introduces a brightness smoothing constraint for the lamp area:

[0086] For example, when the fourth light zone detects that a target will enter the corresponding glare area in the future, the system calculates the new target brightness. The brightness is relatively low; however, since the area was still at a higher brightness level a moment ago, the system will not immediately reduce the brightness abruptly to the target value, but will instead use a smoothing factor. Gradual adjustments are made to gradually reduce the brightness of the light area over multiple control cycles.

[0087] Similarly, when the target leaves the predicted glare area, the system will not immediately restore the maximum brightness, but will gradually restore the high beam output according to the smoothing constraint, thereby avoiding obvious brightness jumps.

[0088] Furthermore, in cases of short-term sensor mismatch, partial target occlusion, or FoV boundary switching, the system can also combine the fused covariance matrix and target acquisition status to adjust the smoothing factor. Make dynamic adjustments: (1) When the target trajectory is stable and the fusion reliability is high, reduce Improve the response speed of the lighting area; (2) When target detection is jittery or fusion is unstable, increase This enhances the stability of brightness changes.

[0089] Therefore, this embodiment can further improve the continuity of light output and visual comfort on the basis of predictive anti-glare control, avoid frequent flickering of matrix light areas, and improve the night driving experience.

[0090] Step 1 introduces multi-source sensing to reduce the risk of failure of a single sensor in scenarios such as rain, fog, and low light. Step 2 quantifies the sensing reliability of different spatial regions to provide a basis for regional differentiation in subsequent control. Step 4 shifts from responding to the current moment to predicting the future moment and identifying risks in advance. Steps 5 and 6 handle anti-glare and supplementary lighting in a unified framework to avoid light pattern fragmentation. Steps 4 to 6 form a prediction-evaluation-adjustment closed loop to achieve advance light pattern adjustment.

[0091] Based on the above-described preferred embodiments of the present invention, and through the foregoing description, those skilled in the art can make various changes and modifications without departing from the inventive concept. The technical scope of this invention is not limited to the contents of the specification, but must be determined according to the scope of the claims.

Claims

1. A method for anti-glare supplementary lighting control based on multi-sensor confidence and trajectory prediction, characterized in that, Includes the following steps: S1. Multiple sensors collect environmental information and unify spatial coordinates; S2. Construct a FoV confidence grid, count the number of multi-sensor overlap coverages in each grid, calculate the regional fusion confidence, and quantify the perception reliability in different spaces. S3. Fuse targets from multiple sensors to obtain the current state variables; S4. Predict the target and the future trajectory of the vehicle, and calculate the anti-glare risk within the future time window; S5. Determine the priority area for supplementary lighting based on the future trajectory of the vehicle and the reliable area of ​​the environment, and generate the supplementary lighting area in combination with anti-glare constraints; S6, jointly decides the brightness of each lamp zone, smooths it, and outputs it to the matrix, pixel, and DLP module.

2. The anti-glare supplementary lighting control method based on multi-sensor confidence and trajectory prediction according to claim 1, characterized in that, Spatial coordinate unification is performed in S1, specifically as follows: First, a spatial transformation is performed on the local coordinate systems of different sensors to unify the position measurement benchmark. The position measurement vector in the sensor coordinate system is represented as follows: Its position in the vehicle coordinate system is represented as , in, and These are the components of the measurement position vector in the sensor coordinate system. and These are the components of the target position vector in the vehicle coordinate system. Let be the installation yaw angle of the i-th sensor relative to the vehicle coordinate system. and Let i be the installation position of the i-th sensor relative to the vehicle reference point; Next, rotate the velocity vector to unify the reference direction of motion. The velocity vector is then represented as: Its representation in the vehicle coordinate system is as , in, and The velocity component of the target relative to the sensor. and The velocity component of the target relative to the sensor in the vehicle coordinate system; Finally, all sensor measurements of environmental targets are mapped to the same vehicle coordinate system to form a unified target observation set.

3. The anti-glare supplementary lighting control method based on multi-sensor confidence and trajectory prediction according to claim 1, characterized in that, The FoV confidence grid is constructed in S2, specifically including: First, define the modeling area in front of and around the vehicle as follows: , Where x and y are the positions of a point in the grid coordinate system. and Boundary values ​​for mesh modeling; Set the grid vertical resolution to The horizontal resolution of the grid is Calculate any point Corresponding two-dimensional grid coordinates The calculation formula is: , Where floor() is the floor function; Then use the one-dimensional index conversion formula Linearize the two-dimensional grid into a unique index, where, Vertical grid index , Horizontal grid index , Total number of vertical grids ; Finally, a grid structure with fast table lookup was constructed to record the sensor coverage, overlap, regional confidence level, and historical control status of each spatial unit.

4. The anti-glare supplementary lighting control method based on multi-sensor confidence and trajectory prediction according to claim 3, characterized in that, S2 calculates the number of multi-sensor overlap coverages for each grid, specifically including: The spatial location of each grid center is calculated from the FoV grid index using the following formula: ， ； Then, the coordinates are transformed to the coordinate systems of each sensor. The calculation formula is as follows: ， in, and The coordinates of the target grid center in the sensor coordinate system. It is the transpose of the rotation matrix. and These represent the sensor's longitudinal and lateral positions, respectively. Calculate the polar coordinates of the grid center in the sensor coordinate system: the distance r of the grid center relative to the sensor and the azimuth deflection angle θ. in, , , If r and θ both fall within the sensor FoV boundary, then the center of the grid is determined to be covered by the sensor. Finally, the number of radar, camera, and modal coverages of each grid is counted to generate a FoV overlap map.

5. The anti-glare supplementary lighting control method based on multi-sensor confidence and trajectory prediction according to claim 4, characterized in that, Based on the FoV overlap plot, a FoV overlap confidence score is introduced, specifically: First, define the region confidence function related to distance and azimuth: Confidence function related to azimuth angle , Distance-related confidence function ; in, Here, is the Sigmoid function, a represents the steepness of the angle decay adjustment, b represents the offset of the Sigmoid function in the angular direction, and c represents the distance decay coefficient. Further define the regional confidence function ; Then calculate the overall FoV overlap confidence for the same grid. The expression is: ， in, The total number of radar sensors. This represents the total number of camera sensors. and These are the confidence functions for the k-th radar and the m-th camera, respectively. and It is an adjustable parameter for Sigmoid angle decay. It is an adjustable parameter for exponential distance decay; and It is an adjustable parameter for Sigmoid angle decay. It is an adjustable parameter for exponential distance decay; Further, a normalized form is introduced, and the expression is: , Finally, the fusion confidence score is obtained by combining sensor health status, target consistency, and scene stability. : , in, , , and To calibrate the weights, It is the confidence level of the sensor's health status. It is the confidence level of scene stability.

6. The anti-glare supplementary lighting control method based on multi-sensor confidence and trajectory prediction according to claim 1, characterized in that, S3 specifically includes: After completing spatial alignment and confidence calculation, the observations of the targets from each sensor are fused. The fused data yields at least the following state variables for each target: , Where x represents the longitudinal position of the target in the vehicle coordinate system, and y represents the lateral position of the target in the vehicle coordinate system. The target's longitudinal relative velocity, The target's lateral relative velocity; At the same time, the corresponding fusion covariance matrix P and fusion confidence are retained for this objective. .

7. The anti-glare supplementary lighting control method based on multi-sensor confidence and trajectory prediction according to claim 1, characterized in that, In S4, the predicted target and the future trajectory of the vehicle are as follows: For the future A constant velocity model is used for first-order prediction, and the predicted target state is: ， ， in, For the goal of the future moment The predicted vertical position, For the goal of the future moment The predicted lateral position, Let the target be the vertical position at time t. Let the target's horizontal position be at time t. For the prediction time interval; The future lighting center trajectory of the vehicle is characterized by combining the vehicle's curvature prediction, and the expression is: ， in, This represents the longitudinal path independent variable in the road coordinate system / vehicle coordinate system. This represents the lateral offset at the vertical position x. , , and These are the fitting parameters obtained based on lane lines, road curvature, and vehicle motion state. Finally, the future fill light center line and the boundary of the high-brightness area of ​​the vehicle headlights are determined along the predicted trajectory.

8. The anti-glare supplementary lighting control method based on multi-sensor confidence and trajectory prediction according to claim 7, characterized in that, Considering the effects of target acceleration or road curvature, future moments The second-order prediction model is used to predict the target state as follows: ， ， in, For the longitudinal acceleration component of the target, The acceleration component of the target in the lateral direction.

9. The anti-glare supplementary lighting control method based on multi-sensor confidence and trajectory prediction according to claim 7, characterized in that, S4 calculates the anti-glare risk within a future time window, specifically including: Define the goal at a future moment Anti-glare risk index for: ， in, To integrate confidence levels for the target future moments. The degree of overlap between the target future location and the high-brightness area of ​​the light. Weights for target type; Define future time windows Comprehensive anti-glare risk within for: , when Greater than the preset threshold In such cases, the corresponding light area is directly shaded or its brightness reduced to achieve predictive anti-glare.

10. The anti-glare supplementary lighting control method based on multi-sensor confidence and trajectory prediction according to claim 1, characterized in that, S5 Mid-Light Priority Area for: ， in: The area representing the vehicle's future driving path; Areas where there are no high-level anti-glare risk targets within the future time window; This represents a high-confidence region where the FoV overlap confidence level is greater than a threshold. In curve scenarios, priority should be given to forward illumination of the inner side of the curve, the road exit direction, and the area near the edge of the lane; in oncoming traffic scenarios, priority should be given to maintaining high beam illumination on the non-glare side of the target.

11. The anti-glare supplementary lighting control method based on multi-sensor confidence and trajectory prediction according to claim 10, characterized in that, S6 specifically refers to: First, the brightness of each light zone is determined jointly based on the predicted anti-glare zone and the supplementary lighting zone. For the i-th light zone, its output brightness is defined. for: ， in, The base brightness of the i-th light zone, This represents the anti-glare suppression amount corresponding to the i-th light zone. This represents the supplementary lighting enhancement amount corresponding to the i-th lighting zone. and These are the weighting coefficients; Apply a smoothing constraint to the brightness of the lamp area at adjacent time intervals: ， in, Let η be the decision brightness of the i-th light zone at time t, and η be the smoothing factor. For the time before the i-th light zone 'output brightness' For the current decision-making brightness, Finally, the headlight controller outputs the brightness of each light zone to the matrix LED module, pixel module, or DLP projection module to achieve predictive anti-glare and supplemental lighting control.

12. A system for implementing the anti-glare supplementary lighting control method based on multi-sensor confidence and trajectory prediction as described in any one of claims 1 to 11, characterized in that, include: The environmental perception module collects information about the environment around the vehicle through a forward-facing camera and a forward-facing millimeter-wave radar. The data preprocessing and spatial alignment module performs spatial unification of sensor coordinates, mapping the target observation data of each sensor to the same vehicle coordinate system to form a unified target observation set; The FoV mesh and overlap confidence calculation module constructs a FoV lookup table-style mesh, counts the number of overlaps of each mesh covered by multiple sensors, and calculates the region fusion confidence based on distance, azimuth angle, sensor health status, and scene stability. The target fusion and trajectory prediction module fuses targets from multiple sensors to obtain target state variables and predicts the future state of the targets and the vehicle's driving trajectory. The anti-glare risk assessment module calculates the anti-glare risk index of the target at future moments, accumulates the comprehensive anti-glare risk within the future time window, and determines whether the target has entered a bright or dazzling area. The supplementary lighting area generation module determines the priority area for supplementary lighting based on the future trajectory of the vehicle and the reliable area of ​​the environment, and generates the supplementary lighting area under the condition of meeting the anti-glare constraint. The headlight control module jointly decides the brightness of each light zone based on the anti-glare zone and the supplementary light zone, and outputs the result to the matrix, pixel or DLP module after smoothing constraint to achieve predictive anti-glare and supplementary light control.

Images