JP7408236B2 - Position estimation method and position estimation device - Google Patents

Description

The present invention relates to a position estimation method and a position estimation device.

Patent Document 1 below discloses that a sensor such as LIDAR (Light Detection and Ranging, Laser Imaging Detection and Ranging) is mounted on the own vehicle, and information detected by the sensor is compared with map information to determine the current position of the own vehicle. A technique for estimating this is described.

International Publication No. 2018/180096 pamphlet

However, if the three-dimensional position information of objects around the host vehicle acquired by the sensor is not sufficient, the accuracy of estimating the position of the host vehicle in the vertical direction and the accuracy of estimating the roll angle and pitch angle may decrease. As a result, the accuracy of estimating the position of the host vehicle in the horizontal direction may also be affected.
An object of the present invention is to reduce the decrease in the accuracy of estimating the position of the own vehicle when the three-dimensional position information of objects around the own vehicle obtained by a sensor is insufficient.

In the position estimation method according to one aspect of the present invention, first position information, which is three-dimensional position information of objects around the own vehicle relative to the own vehicle, is acquired by a sensor mounted on the own vehicle, and the sensor is attached to the own vehicle. Based on the position, second position information is generated that is three-dimensional position information for the own vehicle on the road surface on which the own vehicle is traveling, and known three-dimensional position information on objects around the own vehicle and the road surface on which the own vehicle is traveling is generated. The location of the host vehicle is estimated by comparing the map information, the first location information, and the second location information.

According to one embodiment of the present invention, it is possible to reduce the decrease in the accuracy of estimating the position of the own vehicle when the three-dimensional position information of objects around the own vehicle obtained by the sensor is insufficient.

FIG. 1 is a block diagram of an example of a schematic configuration of a vehicle equipped with a position estimating device according to an embodiment. FIG. 3 is an explanatory diagram of the reason why vertical estimation accuracy decreases in position estimation using three-dimensional position information from a sensor. FIG. 3 is an explanatory diagram of the reason why vertical estimation accuracy decreases in position estimation using three-dimensional position information from a sensor. FIG. 3 is an explanatory diagram of the reason why estimation accuracy in the horizontal direction decreases in position estimation using three-dimensional position information from a sensor. It is an explanatory diagram of an example of the position estimation method of an embodiment. FIG. 1 is a block diagram of an example of a schematic configuration of a position estimation device. FIG. 3 is an explanatory diagram of correction of point cloud data based on the amount of movement of the host vehicle. FIG. 3 is an explanatory diagram of a first example of a setting range of a virtual point group. FIG. 7 is an explanatory diagram of a second example of a setting range of a virtual point group. FIG. 7 is an explanatory diagram of a third example of a setting range of a virtual point group. FIG. 7 is an explanatory diagram of a fourth example of a setting range of a virtual point group. FIG. 7 is an explanatory diagram of a fifth example of a setting range of a virtual point group. FIG. 3 is an explanatory diagram of an example of a process of estimating the position of the host vehicle by comparing point cloud data, virtual point cloud data, and map point cloud data. It is a flow chart of an example of a position estimation method of an embodiment.

Embodiments of the present invention will be described below with reference to the drawings. In the description of the drawings, the same or similar parts are given the same or similar symbols, and overlapping explanations are omitted. Each drawing is schematic and may differ from the actual drawing. The embodiments shown below illustrate devices and methods for embodying the technical idea of the present invention, and the technical idea of the present invention is specific to the devices and methods illustrated in the embodiments below. It's not something you do. The technical idea of the present invention can be modified in various ways within the technical scope described in the claims.

(composition)
Please refer to FIG. Hereinafter, a case will be described in which the current position of a vehicle is estimated as an example of a moving object, but the present invention can be widely applied to estimating the current position of various moving objects, not just vehicles.
The own vehicle 1 is equipped with a position estimation device 10 . The position estimation device 10 includes an object sensor 11, a map database 12, a positioning device 13, a vehicle sensor 14, and a controller 15. In FIG. 1, the map database is expressed as "map DB". The same applies to FIG.

The object sensor 11 is a sensor that detects the position of an object surface around the own vehicle 1 and outputs three-dimensional position information of the detected position.
For example, the object sensor 11 may include a distance measurement sensor such as LIDAR, laser radar, millimeter wave radar, or sonar. Further, the object sensor 11 may be a camera mounted on the own vehicle 1. For example, as the object sensor 11, a sensor that scans objects around the own vehicle 1 in the horizontal direction to obtain three-dimensional position information of the objects may be used.

For example, the object sensor 11 outputs the acquired three-dimensional position information of the object surface around the own vehicle 1 relative to the own vehicle to the controller 15 as three-dimensional point group data. The three-dimensional position information acquired by the object sensor 11 is an example of "first position information" described in the claims. Note that the three-dimensional position information acquired by the object sensor 11 is relative position information based on the position of the object sensor 11 (that is, the position of the own vehicle 1), and is based on the vehicle coordinate system centered on the position of the own vehicle 1. It is expressed as
Note that the three-dimensional position information provided by the object sensor 11 is not necessarily limited to point cloud data, and may be, for example, position information of image feature points or explicit landmarks.

The map database 12 is composed of three-dimensional position information indicating the positions of roads, known objects existing in the vicinity, and the road surface acquired using sensors such as cameras, LIDAR, laser radar, millimeter wave radar, and sonar. This is a database of map information. The map database 12 is stored in a predetermined storage device such as a flash memory or a hard disk drive. The 3D position information that constitutes the map database 12 includes, for example, 3D point cloud data and various forms of information extracted from the 3D point cloud data for position estimation (feature points, targets, landmarks, statistical data, etc.). approximate value).

The three-dimensional position information stored in the map database 12 is expressed in a predetermined processing coordinate system (for example, a map coordinate system or a world coordinate system), and indicates a position on a map. On the other hand, the three-dimensional position information acquired by the object sensor 11 is relative position information based on the position of the object sensor 11 (that is, the position of the own vehicle 1), and is the vehicle coordinate information centered on the position of the own vehicle 1. It is expressed as a system.
Hereinafter, point cloud data that constitutes map information will be referred to as "map point cloud data," and point cloud data acquired by the object sensor 11 will be referred to as "sensor point cloud data."

The positioning device 13 measures the current position of the own vehicle 1. The positioning device 13 may include, for example, a Global Positioning System (GNSS) receiver. The GNSS receiver is, for example, a global positioning system (GPS) receiver or the like, and measures the current position of the host vehicle 1 by receiving radio waves from a plurality of navigation satellites.

The vehicle sensor 14 is mounted on the own vehicle 1 and detects various information (vehicle signals) obtained from the own vehicle 1. The vehicle sensor 14 includes, for example, a vehicle speed sensor that detects the running speed (vehicle speed) of the own vehicle 1, a wheel speed sensor that detects the rotational speed of each tire included in the own vehicle 1, and an acceleration ( A 3-axis acceleration sensor (G sensor) that detects the steering angle (including deceleration), a steering angle sensor that detects the steering angle (including the turning angle), a gyro sensor that detects the angular velocity generated in the own vehicle 1, and a yaw rate that detects the yaw rate. The sensor includes an accelerator sensor that detects the accelerator opening of the host vehicle 1, and a brake sensor that detects the amount of brake operation by the driver.

The controller 15 determines the current three-dimensional position and three-axis orientation (i.e., six degrees of freedom) of the own vehicle 1 based on the sensor point cloud data from the object sensor 11 and the map point cloud data stored in the map database 12. This is an electronic control unit (ECU) that estimates the position and orientation.
Hereinafter, the estimation of the current three-dimensional position and three-axis orientation of the own vehicle 1 by the controller 15 may be simply referred to as "position estimation".

The controller 15 includes a processor 16 and peripheral components such as a storage device 17. The processor 16 may be, for example, a CPU (Central Processing Unit) or an MPU (Micro-Processing Unit).
The storage device 17 may include a semiconductor storage device, a magnetic storage device, an optical storage device, or the like. The storage device 17 may include memories such as a register, a cache memory, a ROM (Read Only Memory) used as a main storage device, and a RAM (Random Access Memory).
The functions of the controller 15 described below are realized, for example, by the processor 16 executing a computer program stored in the storage device 17.

Note that the controller 15 may be formed of dedicated hardware for executing each information process described below.
For example, the controller 15 may include a functional logic circuit set in a general-purpose semiconductor integrated circuit. For example, the controller 15 may include a programmable logic device (PLD) such as a field-programmable gate array (FPGA).

Next, an example of position estimation by the controller 15 will be described.
See FIGS. 2A and 2B. When estimating the position of the own vehicle 1 by comparing the sensor point cloud data 21 acquired by the object sensor 11 and the map point cloud data, the acquired sensor point cloud data 21 is insufficient. The vertical position may become unclear.
For example, LIDAR includes a sensor with a narrow viewing angle (so-called line scanner), and the information obtained by this sensor is point group information distributed in a plane, and information regarding the vertical direction, for example, is degenerated.

Then, for example, as shown in FIGS. 2A and 2B, when sensor point cloud data 21 of an object 20 whose shape changes little in the vertical direction is obtained, this sensor point cloud data 21 and map point cloud data are combined. Even if the comparison is made, the vertical position of the host vehicle 1 becomes unknown, and there is a possibility that the estimation accuracy of the position of the host vehicle 1 in the vertical direction decreases.

As described above, if the accuracy of the vertical position of the own vehicle 1 decreases, there is a risk that the estimation accuracy of the horizontal position will decrease. An example of the reason will be explained with reference to FIG. 2C.
For example, the object surfaces 22 and 23 having different vertical positions have different positions in the distance measurement direction by the object sensor 11. As a result, the accuracy of the vertical position of the vehicle 1 has decreased, and the position of the vehicle 1 is estimated by erroneously comparing the sensor point cloud data 21 on the object surface 22 and the map point cloud data on the object surface 23. Then, the accuracy of estimating the horizontal position also decreases.

Therefore, in the position estimation method of the embodiment, attention is paid to the fact that there is always a road surface below the host vehicle 1 and that the position of the road surface can be estimated based on the attachment position of the object sensor 11 to the host vehicle 1.
See FIG. 3. The controller 15 virtually generates three-dimensional position information 25a, 25b, 25c, and 25d of the road surface 24 on which the host vehicle 1 travels, based on the attachment position of the object sensor 11 to the host vehicle 1. The three-dimensional position information 25a to 25d is an example of "second position information" described in the claims.
Similar to the three-dimensional position information acquired by the object sensor 11, the three-dimensional position information 25a to 25d is relative position information based on the position of the object sensor 11 (that is, the position of the own vehicle 1), and It is expressed in the vehicle coordinate system centered at the location of .

For example, the controller 15 may generate the three-dimensional position information 25a to 25d based on the height H from the grounding point of the own vehicle 1 on the road surface 24 to the object sensor 11.
For example, the controller 15 generates the three-dimensional position information 25a to 25d as three-dimensional point group data representing the three-dimensional position of the road surface 24. Hereinafter, the point cloud data of the road surface 24 generated by the controller 15 will be referred to as "virtual point cloud data."

The controller 15 estimates the position of the own vehicle 1 by comparing the three-dimensional position information acquired by the object sensor 11 and the three-dimensional position information of the road surface 24 generated by the controller 15 with the map information.
For example, the controller 15 estimates the position of the host vehicle 1 by comparing the sensor point cloud data, the virtual point cloud data, and the map point cloud data.

In this way, in the position estimation method of the embodiment, in addition to comparing the three-dimensional position information 21 acquired by the object sensor 11 with map information, the road surface 24 virtually generated based on the mounting position of the object sensor 11 is compared. The position of the own vehicle 1 is estimated by comparing the three-dimensional position information of 25a to 25d with the map information 25a to 25d. This makes it possible to increase the vertical position information to be compared with the map information.
As a result, even if the three-dimensional position information provided by the object sensor 11 is insufficient, it is possible to reduce the decrease in the accuracy of estimating the position of the vehicle 1 in the vertical direction and the accuracy of estimating the roll angle and pitch angle. Further, the decrease in positional accuracy in the horizontal direction due to the decrease in positional accuracy in the vertical direction can also be reduced.

Hereinafter, an example of the function of the position estimation device 10 will be explained in detail. Hereinafter, an example will be described in which the three-dimensional position information obtained by the object sensor 11 and the three-dimensional position information forming the map database 12 are three-dimensional point group data. However, the present invention is not limited to such a form, and as described above, image feature points, targets, landmarks, and statistical approximations can be used as the three-dimensional position information.
See FIG. 4. The controller 15 includes a point cloud data acquisition section 30, a point cloud data storage section 31, a movement amount calculation section 32, a point cloud data correction section 33, a virtual point cloud addition section 34, and a position estimation section 35.

The point cloud data acquisition unit 30 acquires sensor point cloud data 21 from the object sensor 11 . The point cloud data acquisition unit 30 stores the acquired sensor point cloud data 21 in the point cloud data storage unit 31. For example, the point cloud data storage section 31 may be a storage area secured in the storage device 17 shown in FIG. 1. A storage area for the point cloud data storage section 31 may be secured in the storage device in which the map database 12 is stored.
The movement amount calculation unit 32 calculates the movement amount of the host vehicle 1 from the time point tp at which the sensor point group data 21 was acquired in the previous processing cycle to the time point tc at which the sensor point group data 21 was acquired in the current processing cycle. For example, the movement amount calculation unit 32 calculates changes in the position and attitude (direction) of the host vehicle 1 from time tp to time tc as the movement amount of the host vehicle 1.
For example, the movement amount calculation unit 32 calculates the movement amount of the host vehicle 1 based on the positioning result by the positioning device 13 and the odometry based on the vehicle signal from the vehicle sensor 14.

The point cloud data correction unit 33 stores the data acquired in the previous processing cycle and before that (that is, before the time tc at which the sensor point cloud data 21 is acquired in the current processing cycle) and stored in the point cloud data storage unit 31. The sensor point group data 21 is corrected by the movement amount calculated by the movement amount calculation section 32 and stored in the point group data storage section 31.
An example of the correction process of the sensor point cloud data 21 by the point cloud data correction section 33 will be described with reference to FIG. Reference numeral 1 indicates the position of the host vehicle at time tc when sensor point cloud data 21 is acquired in the current processing cycle.

On the other hand, a broken line 40 indicates the position of the own vehicle 1 at the time point tp when the sensor point cloud data 21 was acquired in the previous processing cycle. A circle plot 21 indicates the position indicated by the sensor point cloud data at time tp.
As described above, the sensor point group data is expressed as coordinates on the vehicle coordinate system, that is, relative positions with respect to the own vehicle 1. Therefore, when the host vehicle 1 moves to the position indicated by reference numeral 1 between time tp and time tc, the position indicated by the sensor point group data 21 changes to the position indicated by the rectangular plot 41. The relative position 41 based on the position of the own vehicle 1 at time tc is the same as the relative position 21 based on the position 40 at time tp.
The point cloud data correction unit 33 corrects the sensor point cloud data 21 representing the relative position shown by the square plot 41 into point cloud data representing the relative position shown by the circular plot 21 based on the position of the own vehicle 1 at time tc. do.

See FIG. 4. The virtual point cloud adding unit 34 virtually generates virtual point group data 25a to 25d indicating the position of the road surface 24 on which the host vehicle 1 travels, based on the attachment position of the object sensor 11 to the host vehicle 1. The virtual point cloud data 25a to 25d are added to the point cloud data to be compared with the map point cloud data in estimating the position of the host vehicle 1. The virtual point group data 25a to 25d may be collectively referred to as "virtual point group data 25."
The virtual point group addition unit 34 calculates the vertical coordinates of the virtual point group data 25 based on the height H from the grounding point of the own vehicle 1 on the road surface 24 to the object sensor 11.

The virtual point group adding unit 34 generates one or more virtual point group data 25 in the horizontal plane range 42 shown in FIGS. 6A to 6C and FIG. 7B, for example. For example, the virtual point cloud adding unit 34 may generate the virtual point cloud data 25 so as to be distributed in these ranges 42.
For example, the virtual point group addition unit 34 may appropriately determine the position of the virtual point group data 25 in the horizontal plane to be located within the range 42, and calculate the vertical position based on the height H.

See FIG. 6A. For example, the virtual point cloud adding unit 34 may assume that the road surface 24 under the host vehicle 1 is substantially flat, and may set the range 42 immediately below the host vehicle 1 as the generation range of the virtual point cloud data 25. . In this case, for example, the virtual point cloud adding unit 34 may arrange the virtual point cloud data 25 so as to be distributed at equal intervals in the range 42 directly below the own vehicle 1, or may arrange the virtual point cloud data 25 unevenly. You may. For example, the virtual point group data 25 may be arranged so that the closer it is to the position of the ground contact point of the own vehicle 1, the more dense it is, and the farther away it is, the more sparse it is.
Further, for example, the virtual point cloud adding unit 34 may set a range 42, which is the range immediately below the own vehicle 1 plus a predetermined margin, as the generation range of the virtual point cloud data 25, as shown in FIG. 6B.
In this case, for example, the virtual point cloud adding unit 34 may arrange the virtual point cloud data 25 so as to be distributed at equal intervals in the range 42 to which the margin has been added, or may arrange the virtual point cloud data 25 unevenly. It's okay. For example, the virtual point group data 25 may be arranged so that the closer it is to the position directly below the own vehicle 1, the more dense it is, and the farther away it is, the more sparse it is. Alternatively, the virtual point group data 25 may be arranged so that the closer it is to the ground contact point of the own vehicle 1, the more dense it is, and the farther away it is, the more sparse it is.

Further, as shown in FIG. 6C, the virtual point group data 25 is distributed in a range 42 that extends over the widthwise position range occupied by the host vehicle 1 and extends rearward from the current position of the host vehicle 1, for example. may be generated.
In this case, for example, the virtual point cloud adding unit 34 may arrange the virtual point cloud data 25 so as to be distributed at equal intervals in the range 42 extending backward, or may arrange the virtual point cloud data 25 unevenly. good. For example, the virtual point group data 25 may be arranged so that the closer it is to the position directly below the own vehicle 1, the more dense it is, and the farther away it is, the more sparse it is. Alternatively, the virtual point group data 25 may be arranged so that the closer it is to the ground contact point of the own vehicle 1, the more dense it is, and the farther away it is, the more sparse it is.
Further, for example, the virtual point cloud adding unit 34 generates the virtual point cloud data 25 in the range 42 directly below the own vehicle 1, and the point cloud data correction unit 33 generates the virtual point cloud data 25 based on the amount of movement of the own vehicle 1. By correcting and accumulating, virtual point group data 25 distributed in a range 42 shown in FIG. 6C may be generated.

Further, the virtual point group adding unit 34 may generate virtual point group data 25a and 25b at the position of the contact point of the own vehicle 1 with respect to the road surface 24, as shown in FIG. 7A. The virtual point group adding unit 34 calculates the position of the ground contact portion of the own vehicle 1 based on the attachment position of the object sensor 11 to the own vehicle 1.
Further, the virtual point group adding unit 34 may generate the virtual point group data 25 so that the virtual point group data 25 is distributed in a range 42 occupied by the locus of the position of the contact point of the host vehicle 1 with respect to the road surface 24, as shown in FIG. 7B.
For example, the virtual point cloud addition unit 34 generates virtual point cloud data 25 at the position of the ground contact area of the own vehicle 1, and the point cloud data correction unit 33 corrects the virtual point cloud data 25 based on the amount of movement of the own vehicle 1. The virtual point group data 25 distributed in the range 42 shown in FIG. 7B may be generated by accumulating the virtual point group data 25.

Note that when generating the virtual point cloud data 25, the virtual point cloud adding unit 34 uses the sensor point cloud data 21 acquired by the point cloud data acquisition unit 30 and the sensor point cloud accumulated in the point cloud data storage unit 31. It may be determined whether the data 21 includes point cloud data of the road surface 24 sufficient for estimating the position of the own vehicle 1.
The virtual point cloud addition unit 34 generates (adds) virtual point cloud data 25 when sufficient point cloud data of the road surface 24 is not included in the sensor point cloud data 21, and adds sufficient point cloud data of the road surface 24 to the sensor point cloud data 21. If it is included in the sensor point group data 21, the virtual point group data 25 may not be generated (added), or the virtual point group data 25 to be generated may be reduced. This reduces the virtual point cloud data 25 to be compared with the map point cloud data. As a result, the calculation load can be reduced.

Specifically, the virtual point cloud addition unit 34 calculates the height H from the mounting position of the object sensor 11 to the road surface 24, and selects a predetermined number of sensor point cloud data 21 whose height difference from the height H is less than or equal to a certain value. In the above cases, it may be determined that there is sufficient sensor point group data 21 of the road surface 24.
Alternatively, the range in which the sensor point cloud data 21 has been acquired is divided into a plurality of equally spaced cells, and in each cell it is determined whether the shape indicated by the sensor point cloud data 21 is parallel to the horizontal plane, and the shape indicated by the sensor point cloud data 21 is determined. If there is a certain number or more of cells whose shape is parallel to the horizontal plane, it may be determined that there is sufficient sensor point cloud data 21 of the road surface 24.

Further, the virtual point cloud adding unit 34 may estimate the flatness of the road surface 24 directly below the host vehicle 1 when generating the virtual point cloud data 25.
The virtual point cloud adding unit 34 generates (adds) virtual point cloud data 25 when the flatness is within the allowable range (for example, when the flatness is equal to or higher than a threshold value), and when the flatness is not within the allowable range (for example, when the flatness is (if the degree is less than a threshold value), the virtual point group data 25 may not be generated (added), or the virtual point group data 25 to be generated may be reduced. This reduces the virtual point cloud data 25 to be compared with the map point cloud data.
Therefore, it is possible to prevent the accuracy of position estimation from decreasing due to generation of the virtual point group data 25 at a location where the assumption that the road surface 24 is flat does not hold.

For example, the virtual point cloud addition unit 34 determines whether or not the shape of the road surface 24 directly below the own vehicle 1 can be approximated in a plane, and if it is determined that the shape cannot be approximated in a plane, it does not generate the virtual point cloud data 25 or The point cloud data 25 may be reduced.
Specifically, the virtual point group addition unit 34 may determine whether plane approximation is possible based on whether the step has been climbed over or whether the curvature of the nearby road surface 24 is sufficiently small. The virtual point group addition unit 34 may determine whether or not the step has been climbed based on whether a large change in acceleration has occurred in the vertical direction using an acceleration sensor or the like. The virtual point group addition unit 34 may estimate the curvature of the road surface 24 based on the time series of vertical acceleration and pitch angle.

The position estimation unit 35 estimates the position of the own vehicle 1 by comparing the sensor point cloud data, the virtual point cloud data, and the map point cloud data. An example of position estimation by the position estimation unit 35 will be described below.
The position estimating unit 35 corrects the position of the own vehicle 1 on the processing coordinate system estimated in the previous processing cycle (that is, the position on the map) by the movement amount calculated by the movement amount calculation unit 32, and calculates the position of the own vehicle 1. Determine the current temporary position.

The position estimation unit 35 assumes that the position of the host vehicle 1 on the processing coordinate system is a temporary position, and converts the relative position information indicated by the sensor point group data and the virtual point group data into absolute position information on the processing coordinate system. Convert to
FIG. 8 shows the positions of the sensor point group data 21a, 21b,... and the virtual point group data 25... in the processing coordinate system.
The position estimating unit 35 selects a combination of map point cloud data M1, M2, M3, . . . corresponding to each of the sensor point cloud data 21a, 21b, .

The position estimating unit 35 calculates the error size (distance) D1, D2, D3 between the sensor point cloud data 21a, 21b,... and the virtual point cloud data 25... and the map point cloud data M1, M2, M3... ... is calculated, and the total error S=D1+D2+D3+... is calculated.
The position estimating unit 35 uses numerical analysis to calculate the combination of the map point cloud data M1, M2, M3, etc. that minimizes the sum S, and the position and orientation of the own vehicle 1, and determines the current position of the own vehicle 1 (on the map). position).

Furthermore, the position estimation unit 35 calculates the matching degree Dm between the sensor point cloud data 21a, 21b, ... and the virtual point cloud data 25... and the map point cloud data M1, M2, M3... .
For example, the position estimation unit 35 may calculate a matching degree Dm that is higher as the minimum sum S is smaller. For example, the position estimation unit 35 may calculate the matching degree Dm based on a decreasing function of the minimum sum S. For example, the position estimation unit 35 may calculate the reciprocal of the increasing function of the minimum sum S (for example, the reciprocal of the minimum sum S) as the matching degree Dm.

In reality, the position estimation unit 35 uses existing ICP (iterative closest point), NDT (Normal Distribution Transform), Kalman filter, extended Kalman filter, particle filter, information filter, Newton method, Levenburg-Marquardt method, etc. Based on this, the estimated value of the current position of the own vehicle 1 and the matching degree Dm may be calculated.

The virtual point cloud adding unit 34 may adjust (change) the density of the virtual point cloud data 25 to be generated according to the matching degree Dm, and may adjust (change) the density of the virtual point cloud data 25 to be generated according to the matching degree Dm. The width of the direction range may be adjusted (changed).
For example, the virtual point cloud adding unit 34 may lower the density of the generated virtual point cloud data 25 as the matching degree Dm is lower. Further, the lower the matching degree Dm, the narrower the horizontal range in which the virtual point group data 25 is generated.

This is because if a large number of virtual point group data 25 with a low degree of matching Dm are generated, excessive matching will occur with such virtual point group data 25, making it impossible to perform accurate estimation. Further, the virtual point cloud adding unit 34 may add noise to the virtual point cloud data 25 when the matching degree Dm is lower than a predetermined value, and may increase the noise as the matching degree Dm decreases. Such a method can also suppress excessive matching with the virtual point group data 25 with a low matching degree Dm.
When narrowing the horizontal range in which the virtual point group data 25 is generated, for example, the vehicle width direction range may be narrowed. A typical road surface is formed so that the surface height decreases toward the road edge for water drainage, and the difference in height from the ground contact point increases as you move away from the ground contact point of your vehicle 1 in the vehicle width direction. This is because it increases.

The virtual point cloud addition unit 34 adjusts the density and/or horizontal range of the virtual point cloud data 25 to be generated in the next and subsequent processing cycles according to the matching degree Dm calculated in the current processing cycle (changes in You can let them do it).
Alternatively, the virtual point cloud addition unit 34 adjusts (changes) the density and/or horizontal range to generate the virtual point cloud data 25 again to be used in the current processing cycle, and the position estimation unit 35 performs position estimation. You can try again.
In addition, when reducing the density and/or horizontal range of the virtual point cloud data 25 according to the matching degree Dm, the position estimation unit 35 selects a virtual point cloud used for position estimation from among the generated virtual point cloud data 25. The point cloud data 25 may be selected and the position estimation redone using virtual point cloud data 25 with a smaller density and/or horizontal extent.

(motion)
Next, an example of the position estimation method according to the embodiment will be described with reference to FIG.
In step S1, the movement amount calculation unit 32 calculates the movement amount of the host vehicle 1 from the time point tp at which the sensor point cloud data 21 was acquired in the previous processing cycle to the time point tc at which the sensor point cloud data 21 was acquired in the current processing cycle. calculate.

In step S2, the point cloud data correction unit 33 stores the data acquired in the previous processing cycle and before that (that is, before the time tc at which the sensor point cloud data 21 is acquired in the current processing cycle) into the point cloud data storage unit 31. The stored sensor point group data 21 is corrected by the movement amount calculated by the movement amount calculation section 32 and is stored in the point group data storage section 31.
In step S<b>3 , the point cloud data acquisition unit 30 acquires the sensor point cloud data 21 from the object sensor 11 and stores it in the point cloud data storage unit 31 .

In step S4, the point cloud data correction unit 33 determines whether the sensor point cloud data 21 stored in the point cloud data storage unit 31 includes point cloud data of the road surface 24 sufficient for estimating the position of the host vehicle 1. Determine. If there is sufficient sensor point group data 21 of the road surface 24 (step S4: Y), the process proceeds to step S5 without the point group data correction unit 33 generating virtual point group data 25.
In step S5, the position estimation unit 35 compares the sensor point cloud data 21 and the map point cloud data to estimate the position of the own vehicle 1. Thereafter, the process proceeds to step S10.

On the other hand, if there is not enough sensor point cloud data 21 of the road surface 24 in step S4 (step S4: N), the process proceeds to step S6.
In step S6, the virtual point group addition unit 34 estimates the flatness of the road surface 24 directly below the host vehicle 1, and determines whether the flatness is within an allowable range. If the flatness is not within the allowable range (step S6: N), the point cloud data correction unit 33 does not generate the virtual point group data 25, and the process proceeds to step S5.
In step S5, the position estimation unit 35 compares the sensor point cloud data 21 and the map point cloud data to estimate the position of the own vehicle 1. Thereafter, the process proceeds to step S10.

On the other hand, if the flatness is within the allowable range in step S6 (step S6: Y), the process proceeds to step S7.
In step S7, the virtual point cloud adding unit 34 virtually generates virtual point cloud data 25 indicating the position of the road surface 24 on which the host vehicle 1 travels, based on the attachment position of the object sensor 11 to the host vehicle 1.
In step S8, the position estimating unit 35 compares the sensor point cloud data 21 and the virtual point cloud data 25 with the map point cloud data to estimate the position of the host vehicle 1. Furthermore, the position estimating unit 35 calculates the degree of coincidence (matching degree) Dm between the sensor point cloud data 21 and the virtual point cloud data 25 and the map point cloud data.

In step S9, the virtual point cloud adding unit 34 adjusts the density of the generated virtual point cloud data 25 according to the matching degree Dm. And/or the horizontal range in which the virtual point cloud data 25 is generated is adjusted according to the matching degree Dm. Thereafter, the process proceeds to step S10.
In step S10, the position estimating device 10 determines whether the ignition switch (IGN) of the host vehicle 1 is turned off. If the ignition switch is not turned off (step S10: N), the process returns to step S1. If the ignition switch is turned off (step S10: Y), the process ends.

(Effects of embodiment)
(1) The point cloud data acquisition unit 30 acquires first position information, which is three-dimensional position information of objects around the own vehicle 1 with respect to the own vehicle 1, using the object sensor 11 mounted on the own vehicle 1. The virtual point group addition unit 34 generates second position information that is three-dimensional position information of the road surface 24 on which the own vehicle 1 travels relative to the own vehicle 1 based on the attachment position of the object sensor 11 to the own vehicle 1 . The position estimation unit 35 compares the first position information and the second position information with map information having known three-dimensional position information of objects around the own vehicle 1 and the road surface on which the own vehicle 1 runs, and determines the position of the own vehicle. Estimate the position of 1.

As a result, even if the estimation accuracy of the vehicle's vertical position, roll angle, and pitch angle becomes low when the position of the own vehicle 1 is estimated using only the first position information, the first position information By estimating the position of the host vehicle 1 using the second position information in addition to the second position information, the accuracy of estimating the vertical position, roll angle, and pitch angle can be improved. This makes it possible to reduce deterioration in the estimation accuracy of the vertical position, roll angle, and pitch angle.

(2) The virtual point cloud addition unit 34 stores the position of the ground contact point of the vehicle 1, the range directly below the vehicle 1, the range directly below the vehicle 1 plus a predetermined margin, or the area occupied by the vehicle 1. Second position information of the road surface 24 in the position range in the vehicle width direction is generated.
When the road surface 24 is flat, the second position information of the road surface 24 can be generated with high accuracy within the above range. As a result, the accuracy of estimating the vertical position, roll angle, and pitch angle can be improved.

(3) The movement amount calculation unit 32 obtains the movement amount of the own vehicle 1. The point cloud data correction section 33 corrects the first position information based on the amount of movement and stores it in the point cloud data storage section 31. The position estimating unit 35 compares the first position information and the second position information accumulated in the point cloud data storage unit 31 with map information to estimate the position of the own vehicle 1.
This makes it possible to estimate the position of the own vehicle 1 using the first position information acquired in a plurality of processing cycles, so that the accuracy of estimating the position of the own vehicle 1 can be improved.

(4) The virtual point cloud addition unit 34 determines whether the first position information of the road surface 24 detected by the object sensor 11 is sufficient for estimating the position of the own vehicle 1, and determines whether the first position information of the road surface 24 is is sufficient for estimating the position of the host vehicle 1, the second position information to be compared with the map information is reduced.
Thereby, if the position of the host vehicle 1 can be estimated with sufficient accuracy using the first position information, the generation of the second position information and the comparison between the map information and the second position information can be omitted, thereby reducing the processing load.

(5) The virtual point cloud addition unit 34 estimates the flatness of the road surface, and reduces the second position information to be compared with the map information when the flatness is less than a threshold value.
Thereby, it is possible to prevent a decrease in the accuracy of position estimation based on the second position information generated at a location where the assumption that the road surface 24 is flat does not hold.

(6) The position estimating unit 35 calculates the degree of matching between the first position information, the second position information, and the map information. When the degree of matching is relatively low, the virtual point cloud addition unit 34 makes the range or density of generating the second position information to be compared with the map information smaller than when the degree of matching is relatively high.
It is considered that locations where the degree of matching between the second location information and the map information is low are not suitable for generating the second location information. Therefore, in such locations, by reducing the second location information to be compared with the map information, the degree of matching can be improved and the accuracy of location measurement can be improved.

(7) The object sensor 11 scans objects around the host vehicle 1 in the horizontal direction to obtain first position information.
According to the present invention, even if the object sensor 11 that scans in the horizontal direction is used to obtain first position information in which information regarding the vertical direction is degenerated, the accuracy of estimating the position of the own vehicle 1 in the vertical direction and the roll angle , it is possible to reduce the decrease in pitch angle estimation accuracy. Further, it is possible to reduce the decrease in positional accuracy in the horizontal direction due to the decrease in positional accuracy in the vertical direction.

DESCRIPTION OF SYMBOLS 1... Own vehicle, 10... Position estimation device, 11... Object sensor, 12... Map database, 13... Positioning device, 14... Vehicle sensor, 15... Controller, 16... Processor, 17... Storage device, 30... Point cloud data acquisition Section, 31...Point cloud data storage section, 33...Point cloud data correction section, 34...Virtual point cloud addition section, 35...Position estimation section

Claims (8)

acquiring first position information that is three-dimensional position information of objects around the own vehicle with respect to the own vehicle using a sensor mounted on the own vehicle;
generating second position information that is three-dimensional position information of a point on the road surface on which the own vehicle travels relative to the own vehicle, based on the mounting position of the sensor on the own vehicle;
Comparing map information having known three-dimensional position information of objects around the own vehicle and points on the road surface on which the own vehicle travels with the first position information and the second position information ,
estimating the position of the own vehicle by calculating the position of the own vehicle so that the sum of errors between the map information, the first position information, and the second position information is small ;
A position estimation method characterized by: the position of the ground contact point of the host vehicle, the range directly below the host vehicle, the range directly below the host vehicle plus a predetermined margin, or the position range in the vehicle width direction occupied by the host vehicle; 2. The position estimation method according to claim 1, further comprising generating two position information. Obtaining the amount of movement of the own vehicle;
correcting and accumulating the first position information based on the amount of movement;
estimating the position of the host vehicle by comparing the second position information and the accumulated first position information with the map information;
The position estimation method according to claim 1 or 2, characterized in that: determining whether the first position information of the road surface detected by the sensor is sufficient for estimating the position of the own vehicle;
reducing the second position information to be compared with the map information when the first position information of the road surface is sufficient for estimating the position of the own vehicle;
The position estimation method according to any one of claims 1 to 3, characterized in that: Estimating the flatness of the road surface,
reducing the second location information to be compared with the map information when the flatness is less than a threshold;
The position estimation method according to any one of claims 1 to 4, characterized in that: Calculating the degree of match between the first location information, the second location information, and the map information;
When the matching degree is relatively low, the range or density of generating the second location information to be compared with the map information is made smaller than when the matching degree is relatively high.
The position estimation method according to any one of claims 1 to 5, characterized in that: 7. The position estimation method according to claim 1, wherein the sensor acquires the first position information by horizontally scanning objects around the host vehicle. a sensor that is mounted on the host vehicle and acquires first position information that is three-dimensional position information of objects around the host vehicle with respect to the host vehicle;
Based on the mounting position of the sensor on the host vehicle, second position information that is three-dimensional position information of a point on the road surface on which the host vehicle is traveling relative to the host vehicle is generated, and Map information having known three-dimensional position information of a point on the road surface on which the vehicle is traveling is compared with the first position information and the second position information , and the map information, the first position information, and the second position information are compared. a controller that estimates the position of the own vehicle by calculating the position of the own vehicle such that the sum of errors between the second position information and the second position information is small ;
A position estimation device comprising:

Images