Vehicle headlamp control method and system, storage medium and electronic equipment
By combining deep reinforcement learning and model predictive control, the problem of response lag and insufficient accuracy of automotive headlights in complex scenarios has been solved, achieving fast, accurate, and safe adaptive dimming, thus improving the safety and comfort of nighttime driving.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- CHERY AUTOMOBILE CO LTD
- Filing Date
- 2026-01-29
- Publication Date
- 2026-04-24
AI Technical Summary
Existing automotive headlight control methods are inadequate for effectively handling complex, nonlinear dynamic traffic scenarios, exhibiting issues such as response lag, safety problems, insufficient accuracy, and multi-objective conflicts, thus failing to achieve efficient adaptive dimming.
By combining deep reinforcement learning models and model predictive control models, the system acquires vehicle driving environment and state information, uses deep reinforcement learning models to generate initial dimming commands, and uses model predictive control models for correction and optimization to ensure safety constraints and accuracy. A multi-objective dynamic reward function is constructed to balance safety, comfort, and energy efficiency.
It achieves rapid response and precise dimming, ensuring safety and energy efficiency, significantly improving the adaptive dimming performance of automotive headlights, with a response speed of 85ms, positioning accuracy of ±0.5°, glare avoidance rate of 100%, and energy consumption reduced by 35%.
Smart Images
Figure CN121912876A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of automotive technology, and in particular to a method and system for controlling vehicle headlights, a storage medium, and an electronic device. Background Technology
[0002] Automotive headlights are lighting devices used for nighttime driving, and adaptive dimming technology for automotive headlights is key to improving nighttime driving safety and comfort. Existing automotive headlight control methods (such as PID control and rule-based control) rely on preset thresholds and fixed mathematical models, making it difficult to effectively cope with complex and nonlinear dynamic traffic scenarios, such as sudden changes in the position of oncoming vehicles, dynamic changes in road curvature, and ambient light scattering effects such as rain and fog. This results in response lag (typically greater than 200ms) and glare risk in beam adjustment.
[0003] In recent years, reinforcement learning (RL) agents have been introduced into adaptive vehicle lighting control in hopes of achieving more intelligent decision-making. However, pure RL agent solutions have the following shortcomings:
[0004] 1. Safety issues: In the early stages of training, RL agents may output unsafe actions (such as shooting a beam of light directly at an oncoming vehicle), and there is a lack of real-time safety verification mechanisms.
[0005] 2. Accuracy issues: Existing RL solutions mostly focus on discrete motion space, which makes it difficult to meet the ±0.5° high-precision continuous angle fine adjustment required by automotive headlight stepper motors.
[0006] 3. Conflicting objectives: It is difficult to balance multiple conflicting control objectives such as safety (visibility coverage), comfort (smooth steering), and energy efficiency (low rate of motion change).
[0007] Therefore, there is an urgent need to provide a method for controlling automotive headlights that can achieve adaptive decision-making under complex operating conditions and ensure safe, accurate, and efficient execution. Summary of the Invention
[0008] The purpose of this invention is to overcome the shortcomings of the prior art and provide a vehicle headlight control method and system, storage medium, and electronic device.
[0009] The present invention provides a method for controlling vehicle headlights, comprising: Obtain vehicle driving environment information and vehicle status information; The vehicle driving environment information and the vehicle state information are input into a preset deep reinforcement learning model, and an initial dimming command is output. The initial dimming command is used to adjust the initial horizontal deflection angle of the headlight in the horizontal direction and / or the initial vertical deflection angle in the vertical direction. The initial dimming command is input into a preset model predictive control model, which then corrects and optimizes the initial dimming command based on preset safety constraints and a motor dynamics model to obtain the target dimming command. Adjust the target horizontal and vertical deflection angles of the headlights according to the target dimming command.
[0010] Furthermore, the construction steps of the deep reinforcement learning model include: Based on the dimensions of the environmental state information and the headlight state information, the topology and parameter space of the deep reinforcement learning network are determined. The control model based on physical rules or the basic model predicts the generation of decision trajectories under preset working conditions. The deep reinforcement learning network is pre-trained through supervised learning to obtain an initial deep reinforcement learning model. In a simulated dynamic driving scenario, the initial deep reinforcement learning model is used to explore closed-loop control, and gradient updates are performed based on the multi-objective reward feedback signal after execution to obtain the deep reinforcement learning model.
[0011] Furthermore, the deep reinforcement learning model employs a multi-objective dynamic reward function during training, which is constructed based on a weighted average of at least one of safety, comfort, and energy efficiency indicators.
[0012] Furthermore, the construction steps of the model predictive control model include: Based on the physical characteristic parameters of the headlight, a headlight dynamic state space equation including rotor position, rotational speed and current state is constructed. The glare suppression requirements are mapped to geometric constraints on beam deflection angles, and the physical limits of the actuators are mapped to boundary constraints on motor state quantities. A cost function is constructed to evaluate the dimming response deviation, actuator smoothness, and energy consumption. The geometric constraints and boundary constraints are then transformed into hard constraint boundaries or soft constraint penalty terms for the model predictive control model.
[0013] Further, the process of correcting and optimizing the initial dimming command by the model predictive control model based on preset safety constraints and a motor dynamics model to obtain the target dimming command includes: Based on the motor dynamics model, the impact of the initial dimming command on the headlight beam and the stepper motor is predicted. During the prediction process, the safety constraints are applied, including: beam avoidance of glare area, maximum motor speed limit, and maximum motor acceleration limit; The optimal control sequence is solved to minimize the deviation between the initial dimming command and the safety constraints, thereby obtaining the target dimming command.
[0014] Furthermore, the vehicle driving environment information includes the distance to oncoming vehicles, road curvature, and ambient light intensity, while the vehicle status information includes vehicle speed, vehicle pitch angle, and current headlight status.
[0015] Furthermore, the safety constraints include at least one of glare suppression constraints and actuator physical limit constraints.
[0016] The technical solution of the present invention also provides a vehicle headlight control system, comprising: The acquisition unit is used to acquire vehicle driving environment information and vehicle status information; The initial dimming command output unit inputs the vehicle driving environment information and the vehicle state information into a preset deep reinforcement learning model and outputs an initial dimming command. The initial dimming command is used to adjust the initial horizontal deflection angle of the headlight in the horizontal direction and / or the initial vertical deflection angle in the vertical direction. The target dimming instruction unit is used to input the initial dimming instruction into a preset model predictive control model, and the model predictive control model corrects and optimizes the initial dimming instruction based on preset safety constraints and motor dynamics model to obtain the target dimming instruction. The adjustment unit is used to adjust the target horizontal and vertical deflection angles of the headlights according to the target dimming command.
[0017] The present invention also provides a computer-readable storage medium that stores computer instructions, which, when executed by a computer, are used to perform all steps of the vehicle headlight control method described above.
[0018] The present invention also provides an electronic device, comprising: At least one processor; and, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the vehicle headlight control method as described above.
[0019] The above technical solution has the following beneficial effects: 1. Fast dynamic response: By complementing the model-independent nature of RL and the local optimization capability of MPC, the response lag problem of traditional control and single RL schemes is solved. Experiments show that the response speed of this system can reach 85ms under sudden road conditions (traditional schemes >200ms), significantly improving driving safety.
[0020] 2. High and stable control accuracy: By adopting DDPG continuous motion space and micro-step drive technology, a positioning accuracy of ±0.5° is achieved, and the control stability (positioning error < ±0.8°) can still be maintained under environmental disturbances (such as changes in road slope).
[0021] 3. Reliable safety assurance: The MPC safety correction module applies real-time hard constraints to RL actions, ensuring 100% avoidance of glare areas (the avoidance rate of traditional RL is only 78%), fundamentally solving the safety problems that may occur during the RL exploration period.
[0022] 4. Energy efficiency optimization: By embedding the energy efficiency penalty term into the reward function, the frequent start-stop and overspeed operation of the motor are effectively suppressed, and energy consumption is optimized. Compared with the traditional solution, energy consumption is reduced by about 35%. Attached Figure Description
[0023] The disclosure of this invention will become more readily understood by referring to the accompanying drawings. It should be understood that these drawings are for illustrative purposes only and are not intended to limit the scope of protection of this invention. In the drawings: Figure 1 This is a flowchart illustrating a vehicle headlight control method according to an embodiment of the present invention. Figure 2 This is a flowchart illustrating the implementation steps of a deep reinforcement learning model in one embodiment of the present invention. Figure 3 This is a schematic diagram of a vehicle headlight control system according to an embodiment of the present invention; Figure 4 This is a schematic diagram of the hardware structure of an electronic device for controlling vehicle headlights, provided as an embodiment of the present invention. Detailed Implementation
[0024] The specific embodiments of the present invention will be further described below with reference to the accompanying drawings.
[0025] It is readily understood that, based on the technical solution of this invention, various structural and implementation methods can be interchanged by those skilled in the art without altering the essential spirit of the invention. Therefore, the following detailed embodiments and accompanying drawings are merely illustrative examples of the technical solution of this invention and should not be considered as the entirety of the invention or as limitations or restrictions on the technical solution of the invention.
[0026] The directional terms such as up, down, left, right, front, back, front, back, top, and bottom mentioned or possibly used in this specification are defined relative to the structures shown in the accompanying drawings. They are relative concepts and may therefore vary depending on their location and usage. Therefore, these or other directional terms should not be interpreted as restrictive.
[0027] like Figure 1 As shown, an embodiment of the present invention provides a vehicle headlight control method, comprising: Step S101: Obtain vehicle driving environment information and vehicle status information; Step S102: Input the vehicle driving environment information and the vehicle state information into a preset deep reinforcement learning model, and output an initial dimming command. The initial dimming command is used to adjust the initial horizontal deflection angle of the headlight in the horizontal direction and / or the initial vertical deflection angle in the vertical direction. Step S103: Input the initial dimming command into the preset model predictive control model, and the model predictive control model corrects and optimizes the initial dimming command based on the preset safety constraints and motor dynamics model to obtain the target dimming command; Step S104: Adjust the target horizontal and vertical deflection angles of the headlights according to the target dimming command.
[0028] This invention can be applied to electronic devices with processing capabilities, such as vehicle controllers. For example, the Electronic Control Unit (ECU) of a vehicle.
[0029] Specifically, step S101 is executed first to obtain vehicle driving environment information and vehicle status information.
[0030] In some embodiments, the vehicle driving environment information includes the distance to oncoming vehicles, road curvature, and ambient light intensity, and the vehicle status information includes vehicle speed, vehicle pitch angle, and current headlight status.
[0031] Specifically, the distance to oncoming vehicles is obtained using a lidar sensor positioned at the center of the front of the roof. Road curvature is obtained using a binocular camera located inside the windshield. Ambient light intensity is obtained using a light sensor integrated into the headlight assembly; the light sensor's range is 0-100,000 Lux, with a compensation threshold greater than 15,000 Lux for rain and fog conditions. The vehicle pitch angle is obtained using an inertial measurement unit (IMU) located at the center of the chassis; this pitch angle is used to compensate for the impact of vehicle slope changes on the light beam; the IMU sensor's measurement range is ±90°, with an accuracy of ±0.1°. Preferably, after acquiring the vehicle's driving environment information and vehicle status information, the controller performs spatiotemporal alignment and normalization processing on this information, and then uses a convolutional neural network (CNN) feature extractor to generate a 128-dimensional environment encoding vector.
[0032] Then, step S102 is executed, whereby the vehicle driving environment information and vehicle status information are input into a preset deep reinforcement learning model, and the initial dimming command is output.
[0033] Specifically, the deep reinforcement learning model uses the Deep Deterministic Policy Gradient (DDPG) algorithm to normalize the vehicle driving environment information and vehicle state information before inputting it into the Convolutional Neural Network (CNN) feature extractor to generate a 128-dimensional environment encoding vector. Based on the environment encoding vector and the current headlight state (including brightness and horizontal / vertical slant angle), it generates an initial dimming command in a continuous two-dimensional vector to adjust the initial horizontal slant angle and / or the initial vertical slant angle of the headlights.
[0034] Next, step S103 is executed, where the initial dimming command is input into the pre-trained model predictive control model to predict the impact of the headlight beam on the driver's field of vision within a future time (e.g., 3 seconds). The model predictive control model uses the initial dimming command as the initial solution, applies motor speed limits and beam boundary constraints (to avoid glare), and, based on safety constraints (glare area boundaries, maximum motor speed) and the motor dynamics model (including inertia and electromagnetic torque equations), continuously optimizes the action sequence and dynamically corrects the initial dimming command output by the deep reinforcement learning model to obtain the target dimming command. This ensures that the beam deflection always meets the safety constraints (glare area illumination intensity less than 1000 Lux) and the motor physical limits (angular acceleration less than or equal to 100° / s²). 2 (Spinning speed less than or equal to 500 steps / second) to avoid direct beam of light onto oncoming vehicles.
[0035] Finally, step S104 is executed to convert the target dimming command into a motor pulse signal and adjust the horizontal tilt angle of the headlight reflector. and vertical deflection angle This allows for beam deflection (±15° horizontal / vertical range) to prevent the beam from directly hitting oncoming vehicles. The rotary encoder provides real-time feedback of the actual rotation angle, forming a closed-loop control system.
[0036] Specifically, horizontal deflection angle Adjustment amount: Controls the left and right deflection of the beam (range: -5° to +5°), vertical deflection angle Adjustment amount: Controls the beam pitch angle (range: -3° to +3°).
[0037] In some embodiments, the safety constraints include at least one of glare suppression constraints and actuator physical limit constraints.
[0038] In this embodiment, the adaptive decision-making capability of RL and the real-time security assurance capability of MPC are used to achieve high-precision and high-safety multi-condition adaptive dimming of the headlight.
[0039] In one embodiment, the construction steps of the deep reinforcement learning model include: Based on the dimensions of the environmental state information and the headlight state information, the topology and parameter space of the deep reinforcement learning network are determined. The control model based on physical rules or the basic model predicts the generation of decision trajectories under preset working conditions. The deep reinforcement learning network is pre-trained through supervised learning to obtain an initial deep reinforcement learning model. In a simulated dynamic driving scenario, the initial deep reinforcement learning model is used to explore closed-loop control, and gradient updates are performed based on the multi-objective reward feedback signal after execution to obtain the deep reinforcement learning model.
[0040] Specifically, firstly, the topology and parameter space of the deep reinforcement learning network are determined based on the dimensions of environmental state information and headlight state information. Then, a control model based on physical rules or a basic model is used to predict the decision trajectory generated by the control model under preset operating conditions. The deep reinforcement learning network is pre-trained through supervised learning to achieve adaptive evolution of the control strategy, resulting in an initial deep reinforcement learning model. Finally, in a simulated dynamic driving scenario, the initial deep reinforcement learning model is used to explore closed-loop control. During the exploration process, the original dimming command output by the deep reinforcement learning model is monitored in real time. If the original dimming command does not meet the preset safety constraints, a penalty term is introduced into the reward feedback to force the deep reinforcement learning model to converge to the region that meets the safety constraints within the search space. The gradient is updated based on the multi-objective reward feedback signal after execution, resulting in the deep reinforcement learning model.
[0041] In some embodiments, the deep reinforcement learning model employs a multi-objective dynamic reward function during training, the reward function being constructed based on a weighted average of at least one of safety, comfort, and energy efficiency indicators.
[0042] In one embodiment, such as Figure 2 As shown, the deep reinforcement learning model is implemented using the following steps: Step S201: Define the state space; State vector It contains 12-dimensional features, divided into three categories: 1. Environmental conditions (4D): distance to oncoming vehicles, road curvature, ambient light intensity, vehicle pitch angle; 2. Motor Status (4D): Horizontal / Vertical Actual Rotation Angle / Horizontal / Vertical Rotation Speed / ; 3. Electrical parameters (4-dimensional): Direct-axis / Quadrature-axis current / Electromagnetic torque .
[0043] Step S202: Motion space design; Action vectors Output two-dimensional continuous values to control the stepper motor's rotation angle adjustment: .
[0044] And constrain the angular acceleration to be less than or equal to 100° / s². 2 To prevent mechanical impact, the rotation speed should be less than or equal to 500 steps / second to avoid losing steps.
[0045] Step S203: The policy network generates the initial dimming command; Using the DDPG algorithm, the Actor network receives a 12-dimensional state vector. Output a 2D continuous action vector .
[0046] Step S204: Training of multi-objective dynamic reward function.
[0047] The policy network is trained using the following reward function to balance safety, comfort, and energy efficiency: , in, For security weights, For comfort weighting, The energy efficiency weight is the energy efficiency factor. To reward beam coverage and punish glare, To reward the smooth change of the beam, This is to punish the motor for frequent start-stop cycles.
[0048] The reward function is defined as follows: 1. Safety Rewards: Positive bonus: The beam covers the effective field of view area in front (coverage > 90%).
[0049] Negative reward: The beam of light shines into the eye area of the driver of an oncoming vehicle (triggers glare detection).
[0050] 2. Comfort Rewards: The beam of light changes smoothly with the curvature of the road (angular acceleration <100° / s²).
[0051] Brightness adaptive: Increases brightness in rainy or foggy conditions and decreases brightness in urban lighting conditions.
[0052] 3. Energy efficiency penalty: The motor frequently starts and stops (rate of change of motion > 10Hz) or runs at overspeed.
[0053] Safety weighting in oncoming traffic scenarios It automatically rises to ensure the beam avoids the safe zone.
[0054] Preferably, Set it to 0.5. Set it to 0.3. Set it to 0.2.
[0055] In one embodiment, to facilitate prediction and improve correction accuracy, the construction steps of the model predictive control model include: Based on the physical characteristic parameters of the headlight, a headlight dynamic state space equation including rotor position, rotational speed and current state is constructed. The glare suppression requirements are mapped to geometric constraints on beam deflection angles, and the physical limits of the actuators are mapped to boundary constraints on motor state quantities. A cost function is constructed to evaluate the dimming response deviation, actuator smoothness, and energy consumption. The geometric constraints and boundary constraints are then transformed into hard constraint boundaries or soft constraint penalty terms for the model predictive control model.
[0056] The vehicle headlight control method described in this application has shown significant advantages in real-world road tests: 1. Response speed: Response latency <100ms (traditional solutions >200ms); 2. Positioning accuracy: ±0.5°; 3. Safety: Glare avoidance rate reaches 100%, and the number of times the motor speed exceeds the limit is reduced to 0; 4. Energy efficiency: Energy consumption is reduced by 35% compared to traditional solutions in scenarios where vehicles meet each other.
[0057] like Figure 3 As shown, the technical solution of the present invention also provides a vehicle headlight control system, comprising: The acquisition unit 301 is used to acquire vehicle driving environment information and vehicle status information; The initial dimming command output unit 302 inputs the vehicle driving environment information and the vehicle state information into a preset deep reinforcement learning model and outputs an initial dimming command. The initial dimming command is used to adjust the initial horizontal deflection angle of the headlight in the horizontal direction and / or the initial vertical deflection angle in the vertical direction. The target dimming instruction unit 303 is used to input the initial dimming instruction into a preset model predictive control model, and the model predictive control model corrects and optimizes the initial dimming instruction based on preset safety constraints and motor dynamics model to obtain the target dimming instruction; The adjustment unit 304 is used to adjust the target horizontal and vertical deflection angles of the headlight according to the target dimming command.
[0058] One embodiment of the present invention provides a computer-readable storage medium for storing computer instructions, which, when executed by a computer, are used to perform all steps of the vehicle headlight control method as described in any of the above method embodiments.
[0059] like Figure 4 As shown, a hardware structure diagram of an electronic device for controlling vehicle headlights according to an embodiment of the present invention is provided, including: At least one processor 401; and, Memory 402 is communicatively connected to at least one processor 401; wherein, The memory 402 stores instructions that can be executed by at least one processor 401, which enables the at least one processor 401 to perform the vehicle headlight control method as described in any of the above method embodiments.
[0060] Figure 4 Take a processor 401 as an example.
[0061] The electronic device is preferably an electronic control unit (ECU).
[0062] The electronic device may also include an input device 403 and an output device 404.
[0063] The processor 401, memory 402, input device 403 and output device 404 can be connected by a bus or other means. The figure shows an example of connection by bus.
[0064] Memory 402, as a non-volatile computer-readable storage medium, can be used to obtain non-volatile software programs, non-volatile computer-executable programs, and modules, such as the program instructions / modules corresponding to the vehicle headlight control method in the embodiments of this application, for example, Figure 1-2 The method flow is shown. The processor 401 executes various functional applications and data processing by running non-volatile software programs, instructions, and modules acquired in the memory 402, thereby implementing the vehicle headlight control method in the above embodiment.
[0065] Memory 402 may include a program acquisition area and a data acquisition area, wherein the program acquisition area may acquire the operating system and applications required for at least one function; the data acquisition area may acquire data created based on the use of the vehicle headlight control method, etc. Furthermore, memory 402 may include high-speed random access memory and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other non-volatile solid-state storage device. In some embodiments, memory 402 may optionally include memory remotely located relative to processor 401, and these remote memories may be connected via a network to the apparatus performing the vehicle headlight control method. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.
[0066] The input device 403 can receive user clicks and generate signal inputs related to user settings and function control of the vehicle headlight control method. The output device 404 may include a display device such as a display screen.
[0067] When the one or more modules are accessed in the memory 402 and are run by the one or more processors 401, the vehicle headlight control method in any of the above method embodiments is executed.
[0068] The above-described product can perform the methods provided in the embodiments of this application, and has the corresponding functional modules and beneficial effects for performing the methods. Technical details not described in detail in this embodiment can be found in the methods provided in the embodiments of this application.
[0069] The above embodiments are only used to illustrate the technical solutions of the embodiments of the present invention, and are not intended to limit them. Although the embodiments of the present invention have been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A method for controlling vehicle headlights, characterized in that, include: Obtain vehicle driving environment information and vehicle status information; The vehicle driving environment information and the vehicle state information are input into a preset deep reinforcement learning model, and an initial dimming command is output. The initial dimming command is used to adjust the initial horizontal deflection angle of the headlight in the horizontal direction and / or the initial vertical deflection angle in the vertical direction. The initial dimming command is input into a preset model predictive control model, which then corrects and optimizes the initial dimming command based on preset safety constraints and a motor dynamics model to obtain the target dimming command. Adjust the target horizontal and vertical deflection angles of the headlights according to the target dimming command.
2. The vehicle headlight control method as described in claim 1, characterized in that, The steps for constructing the deep reinforcement learning model include: Based on the dimensions of the environmental state information and the headlight state information, the topology and parameter space of the deep reinforcement learning network are determined. The control model based on physical rules or the basic model predicts the generation of decision trajectories under preset working conditions. The deep reinforcement learning network is pre-trained through supervised learning to obtain an initial deep reinforcement learning model. In a simulated dynamic driving scenario, the initial deep reinforcement learning model is used to explore closed-loop control, and gradient updates are performed based on the multi-objective reward feedback signal after execution to obtain the deep reinforcement learning model.
3. The vehicle headlight control method as described in claim 1 or 2, characterized in that, The deep reinforcement learning model employs a multi-objective dynamic reward function during training, which is constructed based on a weighted average of at least one of safety, comfort, and energy efficiency indicators.
4. The vehicle headlight control method as described in claim 1, characterized in that, The steps for constructing the model predictive control model include: Based on the physical characteristic parameters of the headlight, a headlight dynamic state space equation including rotor position, rotational speed and current state is constructed. The glare suppression requirements are mapped to geometric constraints on beam deflection angles, and the physical limits of the actuators are mapped to boundary constraints on motor state quantities. A cost function is constructed to evaluate the dimming response deviation, actuator smoothness, and energy consumption. The geometric constraints and boundary constraints are then transformed into hard constraint boundaries or soft constraint penalty terms for the model predictive control model.
5. The vehicle headlight control method as described in claim 1, characterized in that, The model predictive control model corrects and optimizes the initial dimming command based on preset safety constraints and a motor dynamics model to obtain the target dimming command, including: Based on the motor dynamics model, the impact of the initial dimming command on the headlight beam and the stepper motor is predicted. During the prediction process, the safety constraints are applied, including: beam avoidance of glare area, maximum motor speed limit, and maximum motor acceleration limit; The optimal control sequence is solved to minimize the deviation between the initial dimming command and the safety constraints, thereby obtaining the target dimming command.
6. The vehicle headlight control method as described in claim 1, characterized in that, The vehicle driving environment information includes the distance to oncoming vehicles, road curvature, and ambient light intensity; the vehicle status information includes vehicle speed, vehicle pitch angle, and current headlight status.
7. The vehicle headlight control method as described in claim 1, characterized in that, The safety constraints include at least one of glare suppression constraints and actuator physical limit constraints.
8. A vehicle headlight control system, characterized in that, include: The acquisition unit is used to acquire vehicle driving environment information and vehicle status information; The initial dimming command output unit inputs the vehicle driving environment information and the vehicle state information into a preset deep reinforcement learning model and outputs an initial dimming command. The initial dimming command is used to adjust the initial horizontal deflection angle of the headlight in the horizontal direction and / or the initial vertical deflection angle in the vertical direction. The target dimming instruction unit is used to input the initial dimming instruction into a preset model predictive control model, and the model predictive control model corrects and optimizes the initial dimming instruction based on preset safety constraints and motor dynamics model to obtain the target dimming instruction. The adjustment unit is used to adjust the target horizontal and vertical deflection angles of the headlights according to the target dimming command.
9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that, when executed by a computer, are used to perform all the steps of the vehicle headlight control method as described in any one of claims 1-7.
10. An electronic device, characterized in that, include: At least one processor; as well as, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the vehicle headlight control method as described in any one of claims 1-7.