Leveling Angle Control System

Abstract

A leveling angle control system (100) for a vehicle headlight (30), comprising: a target leveling angle calculation unit (41) for calculating a target leveling angle θ for a vehicle headlight (30) at prescribed first point on the basis of point information pertaining to a prescribed second point that a vehicle (10) will reach upon travelling for a prescribed number of seconds or a prescribed distance from the prescribed first point; and a leveling angle control unit (42) for controlling the actual leveling angle at the first point such that the actual leveling angle of the vehicle headlight (30) approaches the target leveling angle θ.

Description

Technical Field

[0001] The present disclosure relates to a leveling angle control system.

Background Art

[0002] In recent years, headlights equipped with an auto-leveling function that automatically adjusts the vertical irradiation range according to the longitudinal tilt of a vehicle have become widespread. For example, Patent Document 1 discloses calculating the tilt angle of a vehicle using a gravity sensor and controlling the optical axis of the headlights based on the calculated tilt angle.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] The technique disclosed in Patent Document relates to calculating the current tilt angle of a vehicle and adjusting the optical axis according to the calculated current tilt angle. However, in a location where the tilt angle changes rapidly, such as where there is a sudden change in the road gradient, it has been difficult to make the optical axis follow the rapid change.

[0005] An object of the present disclosure is to appropriately change the optical axis of a vehicle headlight in response to a rapid change in the road tilt angle even in a location where the road tilt angle changes rapidly. <000​​​​​​​​ The system includes a leveling angle control unit that controls the actual leveling angle at the first point so that the actual leveling angle of the vehicle headlight approaches the target leveling angle θ. [Effects of the Invention]

[0007] According to this disclosure, it becomes possible to appropriately adjust the optical axis of the vehicle's headlights in response to sudden changes in the road's incline. [Brief explanation of the drawing]

[0008] [Figure 1] This is a block diagram showing an example of the configuration of a leveling angle control system according to one embodiment of the present disclosure. [Figure 2] This is a schematic diagram illustrating the measurement angle of a vehicle. [Figure 3] This is a schematic diagram illustrating an example of how to acquire information about road surface angle using LiDAR. [Figure 4] This is a schematic diagram illustrating an example of how to acquire information about road surface angle using LiDAR. [Figure 5] This is a schematic diagram showing an example of an image captured by a camera as a vehicle moves uphill. [Figure 6] This is a schematic diagram showing an example of an image captured by a camera as a vehicle moves downhill. [Figure 7] This is a schematic diagram illustrating an example of reinforcement learning. [Figure 8] This flowchart shows an example of the process related to leveling angle control. [Figure 9] This is an example of location data. [Figure 10] This is a flowchart showing an example of a process related to reinforcement learning. [Figure 11] This flowchart shows an example of the process involved in calculating the virtual leveling angle. [Figure 12] This flowchart shows an example of the process related to leveling angle control during reinforcement learning.

Best Mode for Carrying Out the Invention

[0009] Hereinafter, the present invention will be described based on embodiments with reference to the drawings. The same or equivalent components and members shown in each drawing are denoted by the same reference numerals, and redundant descriptions will be omitted as appropriate. In addition, the dimensions of each member shown in the drawings may be different from the actual dimensions of each member for convenience of explanation.

[0010] (Configuration of the leveling angle control system) First, a leveling angle control system according to an embodiment of the present disclosure will be described. The leveling angle control system according to this embodiment is a system that controls the leveling angle at the current driving point based on the location information of a point ahead of the current driving point of the vehicle. FIG. 1 is a block diagram showing an example of the configuration of a leveling angle control system 100 (hereinafter, also simply referred to as "system 100") according to this embodiment.

[0011] System 100 is a system that controls the leveling angle of the headlamp 30 for a vehicle. System 100 includes, for example, a vehicle 10 and a headlamp 30. The vehicle 10 includes, for example, a sensor unit 11 and a vehicle control unit 16. Note that the sensor unit 11 may be provided in the headlamp 30.

[0012] The sensor unit 11 includes, for example, a camera 12, a LiDAR (Light Detection And Ranging) 13, an acceleration sensor 14, and a position sensor 15. The camera 12 is provided so as to be able to image at least the front of the vehicle 10. The LiDAR 13 is provided so as to be able to acquire an image of at least the front of the vehicle 10. The data obtained by the camera 12 and the LiDAR 13 is output to, for example, an image processing unit 17.

[0013] The acceleration sensor 14 is, for example, a three-axis acceleration sensor that detects accelerations in the directions of the x-axis, y-axis, and z-axis that are orthogonal to each other. The acceleration sensor 14 is attached to the vehicle 10, for example, such that the x-axis follows the longitudinal axis of the vehicle 10, the y-axis follows the lateral axis of the vehicle 10, and the z-axis follows the vertical axis of the vehicle 10.

[0014] Based on the value measured by the acceleration sensor 14, the measured angle φ, which is the inclination angle of the vehicle 10 with respect to the horizontal plane, can be calculated. The measured angle φ is used, for example, in the reinforcement learning of the learning model 52 described later. Further, the measured angle φ may be stored in the storage unit 50 in association with the position information and used in the calculation of the target leveling angle θ by the target leveling angle calculation unit 41 described later.

[0015] FIG. 2 is a schematic diagram for explaining the measured angle φ of the vehicle. The measured angle φ is the sum of the road surface angle θr, which is the inclination angle of the road surface with respect to the horizontal plane, and the vehicle angle θv, which is the inclination angle of the vehicle 10 with respect to the road surface. The acceleration sensor 14 detects, for example, the vector Gx, which is the detected value in the x-axis direction of the gravitational acceleration vector G, and the vector Gz, which is the detected value in the z-axis direction of the gravitational acceleration vector G, and calculates the measured angle φ using the following formula (1). Note that the calculation of the measured angle φ is not limited to the above example, and other known methods may be used. Further, the calculation of the measured angle φ may be executed in the vehicle control unit 16 or the lamp control unit 40 described later based on the data detected by the acceleration sensor 14.

Equation

[0016] Returning to the description of FIG. 1. The position sensor 15 is a sensor that acquires the position information of the vehicle 10 and is, for example, a GPS (Global Positioning System) sensor or a GNSS (Global Navigation Satellite System) sensor. The position information of the vehicle 10 is stored, for example, as part of the location data 51 in the storage unit 50.

[0017] The vehicle control unit 16 controls various operations of the vehicle 10, such as driving. The vehicle control unit 16 includes a processor such as an ASIC (Application Specific Integrated Circuit), FPGA (Field programmable Gate Array), or general-purpose CPU (Central Processing Unit). Although not shown in the diagram, the vehicle 10 also includes, for example, a ROM (Read Only Memory) in which various vehicle control programs are stored, and a RAM (Random Access Memory) in which various vehicle control data is temporarily stored. The processor of the vehicle control unit 16 can load data specified from the various vehicle control programs stored in the ROM onto the RAM and control various operations of the vehicle 10 in cooperation with the RAM.

[0018] In this embodiment, the vehicle control unit 16 functions as an image processing unit 17. As will be described in detail later, the image processing unit 17 can identify location information of a point that will be reached by traveling a predetermined number of seconds or a predetermined distance from the vehicle 10's current location, based on data output from the camera 12 or LiDAR 13.

[0019] The headlight 30 is a light fixture mounted on the vehicle 10 that illuminates the area in front of the vehicle 10. The headlight 30 includes, for example, a light fixture control unit 40, a memory unit 50, and a leveling actuator 60. The light fixture control unit 40 includes, for example, a processor such as an ASIC, FPGA, or general-purpose CPU. The memory unit 50 is composed of, for example, ROM or RAM. The processor of the light fixture control unit 40 loads data specified from a program stored in ROM onto RAM and can control various operations of the headlight 30 in cooperation with RAM. The memory unit 50 may be located in the vehicle 10, or it may be configured to be located outside the vehicle 10 (for example, in a data center that can communicate with the vehicle 10).

[0020] In this embodiment, the lighting control unit 40 functions, for example, as a target leveling angle calculation unit 41, a leveling angle control unit 42, a road surface angle information acquisition unit 43, and a learning processing unit 44 by reading a program stored in the storage unit 50.

[0021] The target leveling angle calculation unit 41 calculates the target leveling angle θ of the headlight 30 at a predetermined first point based on location information of a predetermined second point reached by the vehicle 10 traveling a predetermined number of seconds (e.g., 1 second) or a predetermined distance (e.g., 10 m) from the first point. Alternatively, the target leveling angle calculation unit 41 may calculate the target leveling angle θ based on a learning model 52, described later, obtained by reinforcement learning based on the location information. "Location information" includes topographic information of that point and various types of information stored in association with the location information of that point (e.g., location data 51, described later). "Location information" may include, for example, the measurement angle φ, "information regarding the road surface angle θr," described later, and the reference leveling angle.

[0022] The leveling angle control unit 42 controls the actual leveling angle of the headlight 30 at the first point so that the actual leveling angle at the first point approaches the target leveling angle θ. The leveling angle control unit 42 performs the control of the actual leveling angle via the leveling actuator 60.

[0023] The road surface angle information acquisition unit 43 acquires information regarding the road surface angle θr at the second point. The "information regarding the road surface angle θr" is not particularly limited, but it is preferably information indicating whether the road surface is uphill or downhill, or information indicating the road surface angle θr. This information can be acquired, for example, using the camera 12 or LiDAR 13.

[0024] Here, we will explain how to acquire information about the road surface angle θr using Figures 3 to 6. Figures 3 and 4 are schematic diagrams showing an example of how to acquire information about the road surface angle θr using LiDAR 13. In the example in Figure 3, the vehicle 10 is heading uphill. In this case, light (for example, light L3) emitted downward from the horizontal axis H of LiDAR 13 will always hit the ground E and be reflected. That is, if the area in front of the vehicle 10 is uphill, LiDAR 13 can detect the reflected light of all light emitted downward from the horizontal axis H. Therefore, if LiDAR 13 detects the reflected light of all light emitted downward from the horizontal axis H, the road surface angle information acquisition unit 43 may be configured to determine that the second point is on an uphill slope. Note that the horizontal axis H is an axis parallel to the horizontal plane.

[0025] Furthermore, in the example shown in Figure 3, a portion of the light emitted upward from the horizontal axis H of the LiDAR 13 (for example, light L2) hits the ground E and is reflected, while another portion (for example, light L1) does not hit the ground E. In other words, when the area in front of the vehicle 10 is on an uphill slope, the LiDAR 13 detects only a portion of the reflected light emitted upward from the horizontal axis H. Therefore, if the LiDAR 13 detects only a portion of the reflected light emitted upward from the horizontal axis H, the road surface angle information acquisition unit 43 may be configured to determine that the second point is on an uphill slope.

[0026] In the example shown in Figure 4, vehicle 10 is moving downhill. In this case, light emitted upward from the horizontal axis H (for example, light L4) does not hit the ground E. That is, if the area in front of vehicle 10 is downhill, LiDAR 13 will not detect reflected light emitted upward from the horizontal axis H. Therefore, if LiDAR 13 does not detect reflected light emitted upward from the horizontal axis H, the road surface angle information acquisition unit 43 may be configured to determine that the second point is on a downhill slope.

[0027] Furthermore, in the example in Figure 4, a portion of the light irradiated downward from the horizontal axis H (for example, light L6) hits the ground E and is reflected, while another portion (for example, light L5) does not hit the ground E. That is, when the area in front of the vehicle 10 is on a downward slope, the LiDAR 13 detects only a portion of the reflected light of the light irradiated downward from the horizontal axis H. Therefore, if the LiDAR 13 detects only a portion of the reflected light of the light irradiated downward from the horizontal axis H, the road surface angle information acquisition unit 43 may be configured to determine that the second point is on a downward slope.

[0028] In the examples shown in Figures 3 and 4, the road surface angle information acquisition unit 43 may calculate the road surface angle θr at the second point based on the three-dimensional image obtained by the LiDAR 13. Conventional image analysis methods can be used without any particular limitations to calculate the road surface angle θr.

[0029] Next, we will explain how to obtain information about the road surface angle θr using camera 12. Figure 5 is a schematic diagram showing an example of an image captured by camera 12 when vehicle 10 is heading uphill. In the example in Figure 5, the image acquired by camera 12 shows road markings that define the vehicle 10's lane: a white or orange left line LL extending in the longitudinal direction to the left of vehicle 10, and a white or orange right line RL extending in the longitudinal direction to the right of vehicle 10.

[0030] When acquiring information about the road surface angle θr from an image captured by the camera 12, for example, the road surface angle information acquisition unit 43 uses image processing such as the Hough transform to identify the left line LL and the right line RL. Next, the road surface angle information acquisition unit 43 determines whether at least one of the left line LL and the right line RL is bent. If at least one is bent, the road surface angle information acquisition unit 43 identifies a first vanishing point where the extension of one line closer to the vehicle 10 than the bend point intersects with the extension of the other line, and a second vanishing point where the extension of one line further from the vehicle 10 than the bend point intersects with the extension of the other line. Then, if the angle between the left line LL and the right line RL with the first vanishing point as its vertex is greater than the angle between the left line LL and the right line RL with the second vanishing point as its vertex, the road surface angle information acquisition unit 43 determines that the second point is on an uphill slope.

[0031] In the example in Figure 5, both the left line LL and the right line RL are bent at line segment X. In this case, the first vanishing point is P1, the intersection of the extension of the portion of the left line LL closer to the vehicle 10 than line segment X and the extension of the portion of the right line RL closer to the vehicle 10 than line segment X. Similarly, the second vanishing point is P2, the intersection of the extension of the portion of the left line LL further from the vehicle 10 than line segment X and the extension of the right line RL further from the vehicle 10 than line segment X. The road surface angle information acquisition unit 43 then compares the angle A formed by the left line LL and the right line RL with intersection point P1 as its vertex with the angle B formed by the left line LL and the right line RL with intersection point P2 as its vertex. If "angle A > angle B", the road surface angle information acquisition unit 43 determines that the second point is on an uphill slope.

[0032] Figure 6 is a schematic diagram showing an example of an image captured by camera 12 as vehicle 10 is moving downhill. In the example in Figure 6, neither the left line LL nor the right line RL is curved, and the only identifiable vanishing point is the intersection point P3. In this case, where only one vanishing point can be identified, the road surface angle information acquisition unit 43 determines whether a line segment parallel to the horizontal direction (left-right direction) in the image is detected within the angle range formed by the left line LL and the right line RL with the vanishing point as its vertex. In the example in Figure 6, a line segment C is detected within the above range, in which case the road surface angle information acquisition unit 43 determines that the second point is on a downhill slope.

[0033] In the examples shown in Figures 5 and 6, the road surface angle information acquisition unit 43 may calculate the road surface angle θr at the second point based on the image captured by the camera 12. For example, in the example of Figure 5, the road surface angle θr may be calculated from the vertical separation distance, taking advantage of the fact that the greater the uphill slope (the larger the road surface angle θr), the greater the vertical separation distance between the first and second vanishing points in the image. The method for calculating the road surface angle θr from the image captured by the camera 12 is not limited to the above examples, and conventionally known methods can be used without particular restriction.

[0034] Furthermore, regarding the method for acquiring the road surface angle θr described above, it is preferable to combine two or more judgment criteria and calculation methods as appropriate, from the viewpoint of improving the accuracy of the acquired information. Also, regarding the method for acquiring the road surface angle θr described above, each process that was described as being performed by the road surface angle information acquisition unit 43 may be performed by the image processing unit 17. In that case, the road surface angle information acquisition unit 43 only needs to acquire the information that has been judged, calculated, etc. by the image processing unit 17.

[0035] Furthermore, information regarding the road surface angle θr may be configured to be calculated based on a machine learning model. In this case, for example, a machine learning model (e.g., deep learning) can be used, which takes images captured by the camera 12 or three-dimensional images acquired by the LiDAR 13 as input and outputs the measured angle φ or road surface angle θr of the second point, calculated based on data detected by the acceleration sensor 14 when the vehicle 10 travels over the second point.

[0036] Returning to the explanation of Figure 1, the learning processing unit 44 performs reinforcement learning on the learning model 52. Reinforcement learning is performed repeatedly, for example, each time the vehicle 10 travels along a predetermined route including the first and second points. The learning processing unit 44 performs reinforcement learning on the learning model 52, for example, setting it so that a larger reward is given the smaller the absolute value of the difference between the target leveling angle θ at the first point and the measured angle φ of the vehicle 10 at the second point, which is measured by the acceleration sensor 14. The learning processing unit 44 performs Q-learning as reinforcement learning, for example, and performs reinforcement learning so that the Q value increases at each point on the predetermined route. Specific examples of Q values ​​will be explained in a later paragraph.

[0037] Figure 7 is a schematic diagram illustrating an example of reinforcement learning. In the example in Figure 7, points N-1, N, and N+1 are points located on the travel route U. Point N is a point reached by vehicle 10 after traveling a predetermined number of seconds or a predetermined distance from point N-1. Similarly, point N+1 is a point reached by vehicle 10 after traveling a predetermined number of seconds or a predetermined distance from point N. Measured angles φ(N-1), φ(N), and φ(N+1) are the measured angles φ at points N-1, N, and N+1, respectively.

[0038] Reinforcement learning is implemented in a system where, for example, a larger reward is given the smaller the absolute difference between the target leveling angle θ at each point and the measured angle φ at the next point, so that the Q value at each point increases. In this reinforcement learning, the closer the target leveling angle θ(N-1) at point N-1 is to the measured angle (φ) at point N, the greater the reward at point N-1. Similarly, the closer the target leveling angle θ(N) at point N is to the measured angle (φ+1) at point N+1, the greater the reward at point N.

[0039] As this reinforcement learning progresses, it becomes possible to control the leveling angle at the current driving point based on the measured angle φ at a point further ahead. Therefore, even if the inclination angle changes abruptly at a point further ahead, the optical axis can be appropriately adjusted in response to the sudden change in inclination angle. As a result, even when the inclination angle changes abruptly, the decrease in forward visibility can be suppressed.

[0040] Returning to the explanation of Figure 1, the learning processing unit 44 may, for example, set a comparison reference value for Q and a reference leveling angle at each of several points on a predetermined driving route, calculate the Q value at each of the several points on the predetermined driving route when the vehicle 10 is driving along the predetermined driving route, and if there is a point where the calculated Q value is greater than the comparison reference value, update the comparison reference value at that point to the calculated Q value, and update the target leveling angle θ used when calculating the Q value as the reference leveling angle at that point, thereby performing reinforcement learning. The comparison reference value and the reference leveling angle are stored in the storage unit 50 as point data 51, for example, associated with the location information of each point. Note that the initial value of the comparison reference value may be the same at each point. Also, the initial value of the reference leveling angle may be the same at each point, or an initial value may not be set.

[0041] The comparison reference value indicates the highest Q value at each point, and the reference leveling angle is the target leveling angle θ at each point when the Q value reaches its highest value. The learning processing unit 44 may, for example, calculate the next target leveling angle θ based on the reference leveling angle. With this configuration, it is expected that the number of learning iterations required for the Q value to converge at each point can be reduced.

[0042] Furthermore, when the vehicle 10 travels along a predetermined route, the learning processing unit 44 may calculate a virtual leveling angle η at each of several points along the predetermined route, and use the virtual leveling angle η as the target leveling angle θ to perform processes related to calculating the Q value, updating the comparison reference value, and updating the reference leveling angle. In this case, it is preferable that the leveling angle control unit 42 does not perform control of the actual leveling angle based on the target leveling angle θ at points where the comparison reference value does not exceed a predetermined threshold. On the other hand, it is preferable that at points where the comparison reference value exceeds a predetermined threshold, the reference leveling angle is used as the target leveling angle θ to perform control of the actual leveling angle. The virtual leveling angle η can be calculated using the same method as the target leveling angle θ.

[0043] Locations with low comparative reference values ​​are locations where it is still difficult to control the leveling angle appropriately. Therefore, in such locations, a virtual leveling angle η is calculated instead of the target leveling angle θ, and reinforcement learning is performed using the virtual leveling angle η, while the actual leveling angle control based on the location information of the previous location is not performed (for example, by configuring it to perform control based on the measured angle φ of the current location as in the conventional method). This prevents, for example, the leveling angle from being changed to an inappropriate value that is randomly selected.

[0044] On the other hand, at points where the reference level is high, a virtual leveling angle η is calculated instead of the target leveling angle θ. By using the virtual leveling angle η for reinforcement learning, the system searches for an even more optimal leveling angle. However, by using the reference leveling angle as the target leveling angle θ, it becomes possible to control the leveling angle more appropriately than with conventional configurations, even when the slope angle changes rapidly at the aforementioned point.

[0045] (Example of operation of a leveling angle control system) Next, an example of the operation of the system 100 according to this embodiment will be described with reference to Figures 8 to 12. Note that the order of each process constituting each flowchart described below may be in any order, as long as no contradictions or inconsistencies arise in the processing content, and may be executed in parallel.

[0046] Figure 8 is a flowchart showing an example of the process related to leveling angle control. In this embodiment, system 100 performs leveling angle control when a predetermined start condition is met, until a predetermined end condition is met.

[0047] The predetermined start and end conditions are not particularly limited, but for example, the start condition may be when the vehicle 10 starts traveling along a predetermined route, and the end condition may be when it finishes traveling along the predetermined route. In this case, the processes from step S2 to step S4 described later will be repeatedly executed along the predetermined route.

[0048] The predetermined driving route may be set by the user of the vehicle 10, or the lighting control unit 40 may set a frequently driven route as the predetermined driving route by referring to the driving history of the vehicle 10. With this configuration, it becomes possible to perform appropriate leveling angle control on routes desired by the user or frequently driven routes. The start and end of driving on the predetermined driving route may be determined, for example, based on position information acquired by the position sensor 15, or based on start and end operations performed by the user of the vehicle 10.

[0049] Furthermore, the predetermined start condition may be that the absolute value of the difference between the road surface angle θr at the first point and the road surface angle θr at the second point is greater than or equal to a predetermined value, and the predetermined end condition may be that the absolute value of the difference between the road surface angle θr at a predetermined point after the first point and the road surface angle θr at the second point relative to that predetermined point is less than a predetermined value. With such a configuration, in places where the change in the road's inclination angle is large, the leveling angle of the headlights 30 can be appropriately changed in response to the change, and in places where the change in the road's inclination angle is small, the load on the lighting control unit 40, etc., can be reduced by controlling the leveling angle as in the conventional method. Note that the system may also be configured to repeatedly execute each process from step S2 to step S4 while the vehicle 10 is in motion without setting predetermined start and end conditions.

[0050] In the example in Figure 8, if the start condition is not met (No in step S1), the system waits until the start condition is met. If the start condition is met (Yes in step S1), steps S2 to S4 are repeatedly executed until the end condition is met.

[0051] In step S2, the lighting control unit 40 calculates the target leveling angle θ at the first point (the current location of the vehicle 10). The target leveling angle θ at the first point is calculated based on the location information of the second point, which the vehicle 10 reaches after traveling a predetermined number of seconds or a predetermined distance from the first point.

[0052] In step S2, for example, the target leveling angle θ can be calculated based on information regarding the road surface angle θr at the second point. Specifically, if the information regarding the road surface angle θr indicates that the second point is on an uphill slope, the lighting control unit 40 may calculate the target leveling angle θ to be greater than the measured angle φ at the first point. Conversely, if the information regarding the road surface angle θr indicates that the second point is on a downhill slope, the lighting control unit 40 may calculate the target leveling angle θ to be less than the measured angle φ at the first point.

[0053] Furthermore, if the information regarding the road surface angle θr indicates the value of the road surface angle θr at the second point, the lighting control unit 40 may calculate the target leveling angle θ based on the road surface angle θr at the second point. In this case, for example, the target leveling angle θ at the first point may be obtained by correcting the road surface angle θr at the second point using the vehicle angle θv at the first point. Also, if the measured angle φ at the second point is stored in the storage unit 50 as point data 51, the lighting control unit 40 may adopt the measured angle φ at the second point as the target leveling angle θ.

[0054] Furthermore, in step S2, the lighting control unit 40 may calculate the target leveling angle θ of the first location based on the learning model 52 obtained by reinforcement learning based on location information.

[0055] Next, in step S3, the lighting control unit 40 controls the leveling angle of the headlight 30 so that the actual leveling angle at the first point approaches the target leveling angle θ. Through the processing in steps S2 and S3, control is performed to achieve a leveling angle appropriate for the slope at the second point, even before arriving at the second point.

[0056] If vehicle 10 has not traveled a predetermined distance from the first point or if a predetermined time has not elapsed since passing the first point (No in step S4), the system waits until the vehicle has traveled the predetermined distance or the predetermined time has elapsed. If vehicle 10 has traveled a predetermined distance from the first point or if a predetermined time has elapsed since passing the first point (Yes in step S4), the system returns to step S2, with the position of vehicle 10 at that time being designated as the first point. The series of processes from step S2 to step S4 are repeatedly executed until the termination condition is met, and the process related to leveling angle control is terminated when the termination condition is met.

[0057] Next, we will explain the reinforcement learning method applied to the learning model 52 by system 100. First, we will explain the overview of reinforcement learning according to this embodiment. Reinforcement learning uses, for example, the action-value function Q, which is represented by equation (2) below. π(s,a) and the state value function V represented by equation (3) below π This can be done using (s).

number

[0058] In equations (2) and (3) above, t is time, s is the current state, s' is the next state, a is the action, and π is the policy indicating what action to take. P and R are the probability of state s transitioning to s' and the reward obtained at that time, respectively. γ is the discount rate for future rewards. Action-value function Q π (s,a) represents the discounted sum of rewards that can be expected to be obtained in the future by taking action a independently of the policy, and then acting according to policy π, given state s. E is the expected value.

[0059] Here, we will explain the location information used in reinforcement learning. Figure 9 shows an example of location data 51. Location data 51 stores the measurement angle φ, comparison reference value, reference leveling angle, and virtual leveling angle η, associated with the location information of each point on a predetermined driving route.

[0060] In this case, the state s(N) at point N can be defined, for example, as follows: State s(N) = Virtual leveling angle η(N) Here, the virtual leveling angle η is a value that can be randomly set within the range of -3° to 2° with respect to the horizontal plane, and has a resolution of 0.1°. Furthermore, it is preferable that the initial value of the virtual leveling angle η be -0.6°. Note that the above range, resolution, and initial value of the virtual leveling angle η are examples, and other values ​​may be used.

[0061] Action a is defined by varying the virtual leveling angle η(N) within the range of -3° to 2°. For example, the following can be adopted as policy π. Point N+1 has a steeper slope than point N: Randomly increase the virtual leveling angle η(N) from the virtual leveling angle η(N-1) up to a maximum of 1°. Point N+1 has a lower slope than point N: The virtual leveling angle η(N) is randomly lowered from the virtual leveling angle η(N-1) up to a maximum of 1°.

[0062] Action a in this strategy can be expressed by the following formula. Action a(N) = Δη(N-1) = Virtual leveling angle η(N) - Virtual leveling angle η(N-1)

[0063] The reward R(N) can be defined, for example, as follows: Reward R(N) = 5° - |Measurement angle φ(N) - Virtual leveling angle η(N-1)| When controlled according to the target value, the reward is maximized at 5°. When deviating most from the target value, the reward becomes zero. Alternatively, the transition probabilities to the next state s may be assumed to be equally likely, or they may be weighted around the initial state, "-0.6° relative to the horizontal plane".

[0064] Under the above premise, for reinforcement learning, for example, Q-learning can be used. Hereinafter, Q π (s,a) is also called the Q-value. When using Q-learning, system 100 optimizes the leveling angle so that the Q-value is always maximized. In doing so, it searches for optimal actions based on past experience and new actions aimed at obtaining rewards, according to the policy π described above. As a result, the learning model 52 learns the state and actions that maximize the Q-value. Then, by replacing the virtual leveling angle η(N) that maximizes the Q-value with the target leveling angle θ(N), the system optimizes the target leveling angle θ.

[0065] The following provides a more detailed explanation of reinforcement learning. Figure 10 is a flowchart illustrating an example of the reinforcement learning process. The reinforcement learning shown in Figure 9 is repeatedly executed when the vehicle 10 travels along a predetermined route. As a result, a learning model 52 capable of appropriate leveling angle control along the predetermined route is obtained.

[0066] First, in step S11, the lighting control unit 40 sets an upper limit on the number of reinforcement learning iterations. The upper limit can be set to, for example, the number of iterations that causes the Q value to exceed a predetermined threshold. Next, in step S12, the lighting control unit 40 detects the start of travel on a predetermined travel route. The predetermined travel route is a specific route that has been set in advance as the target of reinforcement learning. Detection of the predetermined travel route may be performed, for example, based on location information acquired from the position sensor 15, or based on the user's start operation of the vehicle 10.

[0067] Next, in step S13, the lighting control unit 40 starts timing. Timing is performed to determine when to execute reinforcement learning. Subsequently, the series of processes from step S14 to step S21 are repeated until the lighting control unit 40 detects the end of the predetermined travel route.

[0068] In step S14, the lighting control unit 40 determines whether or not it is time for reinforcement learning. In the example in Figure 10, for example, it is determined that it is time for reinforcement learning every predetermined number of seconds (for example, 1 second) that have elapsed since the start of timing in step S13. Alternatively, the timing for reinforcement learning may be every time the vehicle 10 travels a predetermined distance. In this case, in step S13, the measurement of the distance traveled will start instead of timing.

[0069] If it is not the time for reinforcement learning (No in step S14), the system waits until it is the time for reinforcement learning. If it is the time for reinforcement learning (Yes in step S14), in step S15, the lighting control unit 40 determines whether the vehicle 10's speed is equal to or greater than a predetermined speed (for example, 30 km / h). If the vehicle 10's speed is less than the predetermined speed (No in step S15), the system returns to step S14.

[0070] Even in places where the road's incline changes abruptly, if the vehicle 10 is traveling at a slow speed, it is not difficult to make the optical axis of the headlights 30 follow the abrupt change even with a conventional configuration. Therefore, the system 100 according to this embodiment is particularly useful when the vehicle 10 is traveling at a high speed. By configuring the system not to perform reinforcement learning on data when the vehicle 10 is traveling at a speed below a predetermined speed, it is possible to obtain a learning model 52 that is particularly useful when the vehicle 10 is traveling at a high speed.

[0071] If the vehicle 10 is traveling at a predetermined speed or higher (Yes in step S15), in step S16, the lighting control unit 40 calculates a virtual leveling angle η at the first point (current point) based on a predetermined policy. Hereinafter, the first point will also be referred to as point N, the second point reached by the vehicle 10 after traveling a predetermined number of seconds from the first point will be referred to as point N+1, and the point to which the vehicle 10 was traveling a predetermined number of seconds before the first point will be referred to as point N-1.

[0072] Here, we will describe the process of step S16 in detail using Figure 11. Figure 11 is a flowchart showing an example of the process related to the calculation of the virtual leveling angle η, and shows a specific example of the process of step S16. First, in step S31, the lighting control unit 40 acquires information about the road surface angle θr at point N+1. As already explained, information about the road surface angle θr can be acquired based on images captured by the camera 12 or three-dimensional images acquired by the LiDAR 13.

[0073] If, based on the information obtained in step S31, it is determined that point N+1 has a steeper slope than point N (Yes in step S32), then in step S33, the lighting control unit 40 randomly increases the virtual leveling angle η(N) by up to 1° from the initial value or the virtual leveling angle η(N-1). Conversely, if, based on the information obtained in step S31, it is determined that point N+1 has a steeper slope than point N (No in step S32), then in step S34, the lighting control unit 40 randomly decreases the virtual leveling angle η(N) by up to 1° from the initial value or the virtual leveling angle η(N-1) of point N-1. This configuration is expected to reduce the number of learning iterations required to exceed a predetermined threshold for the Q value.

[0074] Furthermore, if the system is configured to acquire the value of the road surface angle θr at point N+1 in step S31, or if the system is configured to acquire the value of the measured angle φ(N+1) at point N+1 in step S31 when the number of times the predetermined driving route has been driven is two or more, the virtual leveling angle η(N) may be calculated based on the value of the road surface angle θr or the value of the measured angle φ(N+1).

[0075] After steps S33 and S34, the process proceeds to step S17 in Figure 10. In step S17, the lighting control unit 40 obtains the measured angle φ(N+1) at point N+1, which was measured upon arrival at point N+1. Next, in step S18, the lighting control unit 40 calculates the Q value at point N based on the virtual leveling angle η(N) and the measured angle φ(N+1).

[0076] If the Q value calculated in step S18 is greater than the comparison reference value for point N included in point data 51 (Yes in step S19), in step S20, the luminaire control unit 40 updates the comparison reference value for point N in point data 51 to the Q value calculated in step S18. Also, in step S21, the luminaire control unit 40 updates the reference leveling angle for point N in point data 51 to the value of the virtual leveling angle η(N) calculated in step S16. On the other hand, if the Q value calculated in step S18 is less than or equal to the comparison reference value for point N included in point data 51 (No in step S19), the process returns to step S14.

[0077] The series of processes from step S14 to step S21 are repeatedly executed along a predetermined travel route. When the end of the predetermined travel route is detected, in step S22, the lighting control unit 40 updates the learning count. If the learning count reaches the upper limit set in step S11 (Yes in step S23), the reinforcement learning process is terminated. If the learning count has not reached the upper limit set in step S11 (No in step S23), the process returns to step S12 and the reinforcement learning process continues.

[0078] When the reinforcement learning process is completed, it becomes possible to calculate the target leveling angle θ based on the obtained learning model 52, and to control the leveling angle based on the target leveling angle θ. Furthermore, even when the reinforcement learning process is ongoing, it is preferable to configure the system to control the leveling angle using the target leveling angle θ calculated based on the learning model 52 at that time, for example, when certain conditions are met.

[0079] Figure 12 is a flowchart showing an example of the process related to leveling angle control during reinforcement learning. Each process shown in the flowchart in Figure 12 is performed in parallel with, for example, the series of processes from step S14 to step S20 in Figure 10.

[0080] First, in step S41, the lighting control unit 40 detects that the vehicle 10 has arrived at point N. Next, in step S42, the lighting control unit 40 reads the comparison reference value at point N from the point data 51 and determines whether the comparison reference value is equal to or greater than a predetermined threshold.

[0081] If the comparison reference value at point N is greater than or equal to a predetermined threshold (Yes in step S42), in step S43, the lighting control unit 40 reads the reference leveling angle at point N from the point data 51 and calculates the target leveling angle θ(N) based on the read reference leveling angle. In step S43, for example, the value of the read reference leveling angle is taken as the target leveling angle θ(N).

[0082] Next, in step S44, the lighting control unit 40 controls the actual leveling angle at point N based on the target leveling angle θ(N) calculated in step S43, and then terminates. On the other hand, if the comparison reference value at point N is less than a predetermined threshold (No in step S42), the processes in steps S43 and S44 are not executed, and the process terminates.

[0083] Even during reinforcement learning, once reinforcement learning has progressed to a certain extent, it is possible to adjust the optical axis of the headlights 30 more appropriately than in the conventional configuration in response to rapid changes in the road gradient. Therefore, it is preferable to use the learning model 52 for leveling angle control. On the other hand, once reinforcement learning has not progressed, the target leveling angle θ calculated by the learning model 52 is not accurate enough. Therefore, it is preferable to limit the machine learning to use only a virtual leveling angle η and not perform actual leveling angle control using the learning model 52.

[0084] Furthermore, the present invention is not limited to the embodiments described above, and can be freely modified and improved as appropriate. In addition, the material, shape, dimensions, numerical values, form, number, and placement of each component in the embodiments described above are arbitrary and not limited as long as they can achieve the present invention.

[0085] For example, the processes described in each flowchart as being performed by the lighting control unit 40 may be performed by the vehicle control unit 16, to the extent that it does not create inconsistencies. Also, each piece of data stored in the storage unit 50 may be stored in the storage unit of the vehicle 10.

[0086] Furthermore, the processing related to reinforcement learning may be performed on a server device that can communicate with the vehicle 10. In this case, for example, the vehicle 10 transmits location information of multiple locations on a predetermined driving route and the measured angle φ of each of the multiple locations to the server device. The server device then performs reinforcement learning on the learning model 52 and transmits the obtained learning model 52 to the vehicle 10, for example, to store it in the storage unit 50.

[0087] This application is based on Japanese Patent Application No. 2021-171946 filed on October 20, 2021, the contents of which are incorporated herein by reference. [Industrial applicability]

[0088] According to this disclosure, it becomes possible to appropriately adjust the optical axis of the vehicle's headlights in response to sudden changes in the road's incline. [Explanation of Symbols]

[0089] 10: Vehicles 11: Sensor section 12: Camera 13: LiDAR 14: Accelerometer 15: Position sensor 16: Vehicle Control Unit 17: Image Processing Unit 30: (Vehicle) Headlights 40: Lighting control unit 41: Target leveling angle calculation unit 42: Leveling Angle Control Unit 43: Road surface angle information acquisition part 44: Learning Processing Unit 50: Storage part 51: Location Data 52: Learning Models 60: Leveling actuator 100: Leveling Angle Control System

Claims

1. A target leveling angle calculation unit calculates the target leveling angle θ of the vehicle's headlights at a predetermined first point based on location information of a predetermined second point reached by the vehicle traveling a predetermined number of seconds or a predetermined distance from the predetermined first point. The system includes a leveling angle control unit that controls the actual leveling angle at the first point so that the actual leveling angle of the vehicle headlight approaches the target leveling angle θ. Leveling angle control system for vehicle headlights.

2. The calculation of the target leveling angle θ and the control of the actual leveling angle are repeatedly performed along a predetermined travel route. The target leveling angle calculation unit calculates the target leveling angle θ at the first point based on a learning model obtained by performing reinforcement learning, which is set to give a larger reward the smaller the absolute value of the difference between the target leveling angle θ at the first point and the measured angle φ of the vehicle at the second point measured by the acceleration sensor. The leveling angle control system according to claim 1.

3. A learning processing unit that performs Q-learning as reinforcement learning for the learning model, The system further includes a storage unit that stores a reference value for the comparison of Q values ​​at each of a plurality of points along the predetermined driving route, The learning processing unit performs reinforcement learning by, when the vehicle travels along the predetermined travel route, calculating the Q value at each of the multiple points along the predetermined travel route, updating the comparison reference value at any point where the calculated Q value is greater than the comparison reference value at that point to the calculated Q value, and storing the target leveling angle θ used to calculate the calculated Q value as a reference leveling angle in association with that point. The target leveling angle calculation unit calculates the target leveling angle θ based on the reference leveling angle. The leveling angle control system according to claim 2.

4. The learning processing unit calculates a virtual leveling angle η at each of the multiple points on the predetermined driving route when the vehicle is traveling along the predetermined driving route, and uses the virtual leveling angle η as the target leveling angle θ to perform processing related to calculating the Q value, updating the comparison reference value, and storing the reference leveling angle. The leveling angle control unit does not perform control of the actual leveling angle based on the target leveling angle θ at points where the comparison reference value does not exceed a predetermined threshold. The leveling angle control system according to claim 3.

5. The aforementioned reinforcement learning is not performed when the vehicle's speed is below a predetermined speed. The leveling angle control system according to claim 2.

6. Furthermore, it includes a road surface angle information acquisition unit that acquires information regarding the road surface angle at the second point, The aforementioned location information includes information regarding the road surface angle, The target leveling angle θ or virtual leveling angle η at the first point is calculated based on the information regarding the road surface angle. A leveling angle control system according to any one of claims 1 to 5.

7. The aforementioned vehicle is equipped with LiDAR, The aforementioned road surface angle information acquisition unit is: If the LiDAR detects reflected light from all the light irradiated downward from the horizontal axis of the LiDAR at the first point, or if the LiDAR detects reflected light from only a portion of the light irradiated upward from the horizontal axis at the first point, then it is determined that the second point is on an uphill slope. If the LiDAR detects only a portion of the reflected light from the light irradiated downward from the horizontal axis at the first point, or if the LiDAR does not detect any reflected light from the light irradiated upward from the horizontal axis at the first point, then it is determined that the second point is on a downward slope. The leveling angle control system according to claim 6.

8. The aforementioned vehicle is equipped with a camera, The road surface angle information acquisition unit identifies a vanishing point in the image acquired by the camera at the first point, and acquires information regarding the road surface angle at the second point based on the vanishing point. The leveling angle control system according to claim 6.

9. The calculation of the target leveling angle θ and the control of the actual leveling angle are performed when the absolute value of the difference between the road surface angle at the first point and the road surface angle at the second point is greater than or equal to a predetermined value. The leveling angle control system according to claim 6.

Images