Output device, control method, program, and storage medium

Abstract

To provide an output device suitable for improving position estimation accuracy, etc. [Solution] When the in-vehicle unit 1 determines that the position estimation accuracy has decreased, it obtains confidence information for voxel data corresponding to voxels around the route from the map DB 10 which contains voxel data. Based on the obtained confidence information, the in-vehicle unit 1 sets a voxel suitable for estimating the vehicle's position as a reference voxel Btag, and outputs control information to the vehicle's electronic control unit to correct the target trajectory of the vehicle so that it passes through a position suitable for measuring the reference voxel Btag.

Description

[Technical Field]

[0001] This invention relates to a technology for controlling vehicles. [Background technology]

[0002] Conventionally, there is a known technique for estimating a vehicle's own position by comparing (matching) shape data of surrounding objects measured using measuring devices such as laser scanners with map information in which the shapes of surrounding objects are stored in advance. For example, Patent Document 1 discloses an autonomous mobile system that determines whether a detected object in a voxel, which is a space divided according to a predetermined rule, is stationary or moving, and performs matching between map information and measurement data for voxels in which stationary objects exist. [Prior art documents] [Patent Documents]

[0003] [Patent Document 1] International release WO2013 / 076829 [Overview of the project] [Problems that the invention aims to solve]

[0004] The accuracy of location information for stationary structures stored in map data varies, and when matching is performed on stationary structures with low-accuracy location information recorded in the map data, the accuracy of the matching using the map data decreases, resulting in a decrease in the accuracy of self-localization. Patent Document 1 does not disclose anything regarding this issue.

[0005] This invention was made to solve the above-mentioned problems, and its main objective is to provide an output device suitable for improving position estimation accuracy and the like. [Means for solving the problem]

[0006] The invention described in claim 1 is an output device comprising: an acquisition unit that acquires at least precision information relating to the position information and weighting information relating to the possibility of occlusion, which are assigned to voxel data in which the position information of an object for each voxel that is a unit area is recorded; an output unit that outputs control information for moving a moving body so as to approach a voxel determined based on the precision information and weighting information acquired by the acquisition unit; and a measurement unit that measures the position of surrounding objects of the moving body, wherein the weighting information is determined based on the height indicated by the position information.

[0007] The invention described in claim 4 is an output device comprising: an acquisition unit that acquires at least precision information relating to the position information, which is assigned to voxel data in which the position information of an object for each voxel that is a unit area is recorded; an output unit that outputs control information for moving a moving body so as to approach a voxel determined based on the precision information acquired by the acquisition unit; and a measurement unit that measures the position of surrounding objects of the moving body, wherein the precision information is determined based on the distance from a measurement vehicle that performs measurements to generate the position information to the object to be measured and the position estimation accuracy of the measurement vehicle.

[0008] The invention described in claim 5 is a control method performed by an output device, comprising: an acquisition step of acquiring at least precision information relating to the position information, which is assigned to voxel data in which the position information of an object for each voxel, which is a unit area, is recorded; an output step of outputting control information for moving a moving body so as to approach a voxel determined based on the precision information acquired in the acquisition step; and a measurement step of measuring the position of surrounding objects of the moving body, wherein the precision information is determined based on the distance from a measuring vehicle that performs measurements to generate the position information to the object to be measured and the position estimation accuracy of the measuring vehicle.

[0009] The invention according to claim 6 is a program executed by a computer, which includes an acquisition unit that acquires at least accuracy information regarding the position information, the accuracy information being attached to voxel data in which position information of an object for each voxel, which is a unit region, is recorded; an output unit that outputs control information for moving a moving object so as to approach a voxel determined based on the accuracy information acquired by the acquisition unit; and a measurement unit that functions as the computer to measure the positions of surrounding objects of the moving object, and the accuracy information is determined based on the distance from a measurement vehicle that performs measurement for generating the position information to the object to be measured and the position estimation accuracy of the measurement vehicle.

Brief Description of the Drawings

[0010] [Figure 1] It is a schematic configuration of a driving support system. [Figure 2] It shows a block configuration of an in-vehicle device. [Figure 3] It shows an example of a schematic data structure of voxel data. [Figure 4] It is a diagram showing the positional relationship between a map-making vehicle equipped with a lidar and an object existing within the measurement range of the map-making vehicle. [Figure 5] It shows a specific example of NDT scan matching. [Figure 6] It shows a specific example of NDT scan matching in which a weighting value is set for each voxel. [Figure 7] It is a flowchart showing a processing procedure executed by an in-vehicle device. [Figure 8] It shows a bird's-eye view of a vehicle when there are ground objects in the left and right front of the road on a three-lane road on one side.

Embodiments for Carrying Out the Invention

[0011] According to a preferred embodiment of the present invention, the output device comprises an acquisition unit that acquires at least precision information relating to the position information, which is attached to map information on which the position information of an object is recorded, and an output unit that outputs control information for controlling a moving object based on the precision information acquired by the acquisition unit. In this embodiment, the output device can suitably control a moving object based on the precision information attached to the map information.

[0012] In one embodiment of the output device described above, the output device further includes a position estimation unit that estimates the position of the moving object by comparing the output of a measurement unit, which measures the position of objects surrounding the moving object, with the position information included in the map information. In this embodiment, the output device can move the moving object to a position where the comparison between the output of the measurement unit and the position information in the map information can be performed with high accuracy, based on the accuracy information attached to the map information, thereby suitably improving the accuracy of the position estimation.

[0013] In another embodiment of the output device described above, the map information includes location information for each region obtained by dividing space according to a predetermined rule, and the position estimation unit weights evaluation values ​​to assess the degree of matching for each region based on the accuracy information. In this embodiment, the output device can suitably improve the position estimation accuracy by increasing the weighting of regions containing highly reliable location information.

[0014] In another embodiment of the output device described above, the map information includes positional information for each region obtained by dividing the space according to a predetermined rule, and the output unit outputs control information for moving the mobile body to a position where the region determined based on the accuracy information is measured by the measurement unit. In this embodiment, the output device can suitably control the mobile body to measure the region determined based on the accuracy information.

[0015] In another embodiment of the output device described above, the output unit outputs control information to move the moving body to the lane closest to the region determined based on the accuracy information, or to move the moving body to the side closer to the region within the lane in which it is traveling. In this embodiment, the output device can control the moving body to reliably and accurately measure the region determined based on the accuracy information.

[0016] In a preferred example of the output device described above, the accuracy information is preferably determined based on the distance from the measuring vehicle that performs the measurement to generate the position information to the object being measured, and the position estimation accuracy of the measuring vehicle.

[0017] In another embodiment of the output device described above, the map information is accompanied by weighting information regarding the possibility of occlusion, and the output unit outputs control information for controlling the moving object based on the accuracy information and the weighting information. In this embodiment, the output device can output control information for controlling the moving object by taking into account both the accuracy information and the weighting information regarding occlusion.

[0018] In another embodiment of the output device described above, the output device further comprises a position estimation unit that estimates the position of the moving object by comparing the output of a measurement unit that measures the position of objects surrounding the moving object with the position information included in the map information, wherein the map information includes position information for each region obtained by dividing space according to a predetermined rule, and the position estimation unit weights an evaluation value that evaluates the degree of matching for each region based on the accuracy information and the weighting information. In this embodiment, the output device can significantly increase the weighting of regions having position information with high reliability based on two viewpoints: occlusion and accuracy, thereby suitably improving the position estimation accuracy.

[0019] In a preferred example of the output device described above, the weighting information is preferably determined based on the height indicated by the position information.

[0020] According to another preferred embodiment of the present invention, the output device comprises an acquisition unit that acquires at least weighting information regarding the possibility of occlusion, which is assigned to map information on which the position information of an object is recorded, and an output unit that outputs control information for controlling a moving object based on the weighting information acquired by the acquisition unit. In this embodiment, the output device can suitably control a moving object based on the weighting information assigned to the map information.

[0021] According to another preferred embodiment of the present invention, a control method performed by an output device comprises: an acquisition step of acquiring at least precision information relating to the position information of an object, which is assigned to map information on which the position information of an object is recorded; and an output step of outputting control information for controlling a moving object based on the precision information acquired in the acquisition step. By performing this control method, the output device can suitably control a moving object based on the precision information assigned to the map information.

[0022] According to another preferred embodiment of the present invention, a computer is configured to function as a program executed by a computer, comprising an acquisition unit that acquires at least precision information relating to location information, which is attached to map information on which the location information of an object is recorded, and an output unit that outputs control information for controlling a moving object based on the precision information acquired by the acquisition unit. By executing this program, the computer can suitably control the moving object based on the precision information attached to the map information. Preferably, the program is stored in a storage medium. [Examples]

[0023] Preferred embodiments of the present invention will be described below with reference to the drawings.

[0024] Figure 1 is a schematic diagram of the driver assistance system according to this embodiment. The driver assistance system shown in Figure 1 includes an on-board unit 1 mounted on the vehicle that performs control related to driver assistance, a Lidar (Light Detection and Ranging, or Laser Illuminated Detection and Ranging) 2, a gyro sensor 3, a vehicle speed sensor 4, and a GPS receiver 5.

[0025] The in-vehicle unit 1 is electrically connected to the lidar 2, gyro sensor 3, vehicle speed sensor 4, and GPS receiver 5, and estimates the position of the vehicle on which the in-vehicle unit 1 is mounted (also called "vehicle position") based on their outputs. Then, based on the estimated vehicle position, the in-vehicle unit 1 performs automatic driving control of the vehicle so that it travels along a set route to a destination. The in-vehicle unit 1 stores a map database (DB: Database) 10 that contains voxel data. Voxel data is data that records position information of stationary structures for each region (also called "voxel") when a 3D space is divided into multiple regions. The voxel data includes data that represents the measured point cloud data of stationary structures within each voxel using a normal distribution, and is used for scan matching using NDT (Normal Distributions Transform), as described later.

[0026] The lidar 2 discretely measures the distance to an object in the external environment by emitting a pulsed laser within a predetermined angular range in the horizontal and vertical directions, and generates three-dimensional point cloud information indicating the position of the object. In this case, the lidar 2 has an irradiation unit that irradiates laser light while changing the irradiation direction, a light receiving unit that receives reflected light (scattered light) of the irradiated laser light, and an output unit that outputs scan data based on the received signal output by the light receiving unit. The scan data is generated based on the irradiation direction corresponding to the laser light received by the light receiving unit and the response delay time of the laser light, which is determined based on the received signal described above. Generally, the accuracy of the lidar's distance measurement is higher the closer the distance to the object, and lower the accuracy the farther the distance. The lidar 2, gyro sensor 3, vehicle speed sensor 4, and GPS receiver 5 each supply output data to the in-vehicle unit 1. The in-vehicle unit 1 is an example of an "output device" in this invention, and the lidar 2 is an example of a "measurement unit" in this invention.

[0027] Figure 2 is a block diagram showing the functional configuration of the in-vehicle unit 1. The in-vehicle unit 1 mainly consists of an interface 11, a storage unit 12, an input unit 14, a control unit 15, and an information output unit 16. Each of these elements is interconnected via a bus line.

[0028] Interface 11 acquires output data from sensors such as the rider 2, gyro sensor 3, vehicle speed sensor 4, and GPS receiver 5, and supplies it to the control unit 15. Interface 11 also supplies signals related to vehicle driving control generated by the control unit 15 to the vehicle's electronic control unit (ECU). The signals transmitted from the control unit 15 to the vehicle's electronic control unit via Interface 11 are an example of "control information" in this invention.

[0029] The storage unit 12 stores programs to be executed by the control unit 15 and information necessary for the control unit 15 to perform predetermined processes. In this embodiment, the storage unit 12 stores a map DB 10 containing voxel data. The map DB 10 may be updated periodically. In this case, for example, the control unit 15 receives partial map information relating to the area to which the vehicle's position belongs from a server device that manages map information via a communication unit (not shown) and reflects it in the map DB 10.

[0030] The input unit 14 includes buttons, a touch panel, a remote controller, a voice input device, etc., for user operation, and accepts inputs such as specifying a destination for route searching and specifying whether to turn autonomous driving on or off. The information output unit 16 includes, for example, a display or speaker that outputs based on the control of the control unit 15.

[0031] The control unit 15 includes a CPU that executes programs and controls the entire in-vehicle unit 1. In this embodiment, the control unit 15 has a vehicle position estimation unit 17 and an automatic driving control unit 18. The control unit 15 is an example of the "acquisition unit," "position estimation unit," "output unit," and "computer" that executes programs in the present invention.

[0032] The vehicle position estimation unit 17 estimates the vehicle's position by performing scan matching based on NDT, using the point cloud data output from the rider 2 and the voxel data corresponding to the voxel to which the point cloud data belongs.

[0033] The automatic driving control unit 18 refers to the map DB 10 and transmits signals necessary for automatic driving control to the vehicle based on the set route and the vehicle's position estimated by the vehicle position estimation unit 17. Based on the set route, the automatic driving control unit 18 sets a target trajectory and controls the vehicle's position by transmitting guide signals to the vehicle so that the vehicle's position estimated by the vehicle position estimation unit 17 is within a predetermined width of the target trajectory. In this embodiment, if the automatic driving control unit 18 determines that the accuracy of the vehicle position estimation by the vehicle position estimation unit 17 has fallen below a predetermined accuracy, it refers to the voxel data and selects a voxel suitable for estimating the vehicle's position. The automatic driving control unit 18 then considers the selected voxel as the voxel to be used as a reference in vehicle position estimation (also called the "reference voxel Btag") and corrects the vehicle's target trajectory so that the vehicle moves to a position suitable for detecting the reference voxel Btag by the lidar 2.

[0034] [Data structure of voxel data] This section describes the voxel data used for scan matching based on NDT. Figure 3 shows an example of the general data structure of the voxel data.

[0035] The voxel data includes parameter information for representing the point cloud within a voxel using a normal distribution. In this embodiment, as shown in Figure 3, it includes voxel ID, voxel coordinates, mean vector, covariance matrix, and confidence information. Here, "voxel coordinates" represent the absolute three-dimensional coordinates of a reference position, such as the center position of each voxel. Each voxel is a cube that divides space into a grid, and its shape and size are predetermined, so the space of each voxel can be identified by its voxel coordinates. The voxel coordinates may also be used as the voxel ID.

[0036] The "mean vector" and "covariance matrix" represent the mean vector and covariance matrix, which correspond to the parameters when representing the point cloud within the target voxel using a normal distribution, and the coordinates of any point "i" within any voxel "k" are

[0037]

number

[0038]

number

[0039]

number

[0040] The "reliability information" includes a first weighting value, which is a weighting value based on the likelihood of occlusion (obstruction by an obstacle), and a second weighting value, which is a weighting value based on the accuracy (precision) of the voxel data (particularly the mean vector and covariance matrix) of the target voxel. In this embodiment, the first weighting value is set to a larger value for voxels that are less likely to be occluded. The first weighting value is an example of the "weighting information" in the present invention.

[0041] The second weighting value is set based, for example, on the accuracy of the map preparation vehicle's self-position estimation when it collects measurement data used to generate voxel data, and the measurement accuracy of the measurement device (e.g., a lidar) that outputs the measurement data. For example, the accuracy is calculated by taking the square root of the sum of the squares of the map preparation vehicle's self-position estimation accuracy and the measurement accuracy of the measurement device, and the reciprocal of the square of that accuracy is taken as the second weighting value. In this embodiment, the second weighting value becomes larger the higher the map preparation vehicle's self-position estimation accuracy when collecting measurement data, and larger the higher the measurement accuracy of the measurement device that outputs the measurement data. Here, if the measurement device is a lidar, the above-mentioned measurement accuracy increases as the distance from the measurement position to the object being measured decreases. Therefore, the second weighting value is determined, for example, based on the distance between the map preparation vehicle's estimated self-position during driving and the position indicated by the voxel coordinates of the target voxel, and the self-position estimation accuracy. Furthermore, in vehicle position estimation methods such as RTK-GPS, it is possible to obtain information regarding the accuracy of the estimated vehicle position in addition to the estimated vehicle position information. The second weighted value is an example of "accuracy information" in this invention.

[0042] Here, we will explain examples of setting the first and second weighting values.

[0043] Figure 4 shows an example of the positional relationship between a map maintenance vehicle equipped with a RIDA and objects within the measurement range of the map maintenance vehicle. In the example in Figure 4, a low-rise building 21, which is a local feature (stationary structure), is located around the road on which the vehicle is traveling, and a high-rise building 22 is located behind the low-rise building 21. In addition, a pedestrian 23 riding a motorcycle is located in front of the low-rise building 21. Here, the map maintenance vehicle is assumed to acquire measurement data representing the position within frames 30-32 using a measuring device such as a RIDA.

[0044] In this case, the portion of the low-rise building 21 enclosed by frame 30 is located at a relatively low position, making it prone to occlusion by moving objects such as pedestrians 23. Therefore, a relatively small first weighting value is set for the voxel data at the location overlapping with frame 30. On the other hand, the portion of the low-rise building 21 enclosed by frame 30 is located relatively close to the map preparation vehicle, so the accuracy of the measurement data acquired by the measurement device is expected to be high. Therefore, a relatively large second weighting value (i.e., a value indicating high accuracy) is set for the voxel data at the location overlapping with frame 30. The portion of the low-rise building 21 enclosed by frame 31 is located at a slightly higher position from the map preparation vehicle, but there remains a possibility of occlusion by tall vehicles such as trucks. Also, because it is located a little further away from the map preparation vehicle, the accuracy of the measurement data acquired by the measurement device is slightly lower than that of the portion enclosed by frame 30. Therefore, a medium first weighting value is set for the voxel data at the location overlapping with frame 31, and a medium second weighting value is set for both.

[0045] On the other hand, the portion of the high-rise building 22 enclosed by frame 32 is located at a relatively high position, making it less susceptible to occlusion caused by moving objects such as pedestrians 23 and other vehicles. Therefore, a relatively large first weighting value is set for the voxel data at the location overlapping with frame 32. Furthermore, since the portion of the high-rise building 22 enclosed by frame 32 is located relatively far from the map maintenance vehicle, the accuracy of the measurement data acquired by measuring devices such as a lidar is expected to be low. Therefore, a relatively small second weighting value (i.e., a value indicating low accuracy) is set for the voxel data at the location overlapping with frame 32.

[0046] Preferably, the first weighting value is set to be larger the higher the target voxel is located relative to the ground surface, and the second weighting value is set to be larger the closer the voxel is to the map preparation vehicle used during measurement.

[0047] [Overview of Scan Matching] Next, scan matching by NDT using voxel data will be described. In this embodiment, as will be described later, the in-vehicle device 1 calculates the value of the evaluation function (evaluation value) obtained by NDT scan matching by weighting it using the reliability information included in the voxel data. Thereby, the in-vehicle device 1 suitably improves the position estimation accuracy based on NDT scan matching.

[0048] Scan matching by NDT assuming a vehicle estimates the following estimation parameter "P" having elements of the movement amount in the road plane (here, xy coordinates) and the orientation of the vehicle.

[0049]

Equation

[0050] Using the above-described estimation parameter P, the coordinates [x k (i), y k (i), z k (i)] T of an arbitrary point in the point cloud data obtained by the lidar 2 are coordinate-transformed, and the transformed coordinates “X′ k ​​​​​​​​​​​​​​​The overall evaluation function "E" (also called the "overall evaluation function") is calculated for all voxels subject to matching as shown by equation (5).

[0052]

number

[0053]

number

[0054] On the other hand, the evaluation function E used in conventional NDT matching is for voxel k. k This is shown by the following equation (6).

[0055]

number

[0056] Subsequently, the in-vehicle unit 1 calculates estimated parameters P that maximize the overall evaluation function E using an arbitrary root-finding algorithm such as Newton's method. Then, the in-vehicle unit 1 applies the estimated parameters P to the vehicle's position predicted from the output of the GPS receiver 5, etc., to estimate the vehicle's position with high accuracy.

[0057] Next, we will explain a specific example of NDT scan matching. For the sake of explanation, we will use a two-dimensional plane as an example below.

[0058] Figure 5(A) shows point clouds measured by a lidar or similar device while driving a map maintenance vehicle in four adjacent voxels "B1" to "B4," indicated by circles. The figure then shows a gradient of the two-dimensional normal distribution created from equations (1) and (2) based on these point clouds. The mean and variance of the normal distribution shown in Figure 5(A) correspond to the mean vector and covariance matrix, respectively, in the voxel data.

[0059] Figure 5(B) shows the point cloud acquired by the RIDA 2 while the vehicle-mounted unit 1 was driving, indicated by stars in Figure 5(A). The positions of the RIDA 2 point cloud, indicated by stars, are aligned with each voxel B1 to B4 based on the estimated position from the output of the GPS receiver 5, etc. In the example in Figure 5(B), there is a discrepancy between the point cloud measured by the map maintenance vehicle (circles) and the point cloud acquired by the vehicle-mounted unit 1 (stars).

[0060] Figure 5(C) shows the state after the point cloud (stars) acquired by the on-board unit 1 has been moved based on the matching results of the NDT scan matching. In Figure 5(C), the parameter P that maximizes the evaluation function E shown in equations (4) and (5) is calculated based on the mean and variance of the normal distribution shown in Figures 5(A) and (B), and the calculated parameter P is applied to the point cloud of stars shown in Figure 5(B). In this case, the discrepancy between the point cloud (circles) measured by the map maintenance vehicle and the point cloud (stars) acquired by the on-board unit 1 is suitably reduced.

[0061] Here, if we calculate the evaluation functions "E1" to "E4" and the overall evaluation function E corresponding to voxels B1 to B4 using the conventionally used general formula (6), these values ​​will be as follows. E1 = 1.3290 E2 = 1.1365 E3 = 1.1100 E4 = 0.9686 E = 4.5441 In this example, there is no significant difference between the evaluation functions E1 to E4 for each voxel, although there are slight differences due to the number of points in the voxel's data pool.

[0062] In this embodiment, a first weighting value and a second weighting value are assigned to each voxel. Therefore, by increasing the weight of a voxel with high confidence, it is possible to improve the degree of matching for that voxel. Below, as an example, a specific example of setting the first weighting value for each voxel will be explained with reference to Figure 6.

[0063] Figure 6(A) shows the matching results when the first weighting values ​​for voxels B1 to B4 are all equal (i.e., the same figure as Figure 5(C)). Figure 6(B) shows the matching results when the first weighting value for voxel B1 is 10 times the weighting value of the other voxels. Figure 6(C) shows the matching results when the first weighting value for voxel B3 is 10 times the weighting value of the other voxels. In all examples, the second weighting values ​​are assumed to be set to equal values.

[0064] In the example in Figure 6(B), the values ​​of the evaluation functions E1 to E4 and the overall evaluation function E corresponding to voxels B1 to B4 are as follows: E1 = 0.3720 E2 = 0.0350 E3 = 0.0379 E4 = 0.0373 E = 0.4823

[0065] As shown in Figure 6(B), matching is performed in such a way that the value of the evaluation function E1 corresponding to voxel B1 is increased, thereby improving the degree of matching in voxel B1. As a result, the discrepancy between the circles and stars in voxel B1 is reduced. Furthermore, although the value of the evaluation function has decreased due to normalization by the number of points in the point cloud, each evaluation function value is in a proportion similar to that of the weighted values.

[0066] Furthermore, in the example in Figure 6(C), the values ​​of the evaluation functions E1 to E4 and the overall evaluation function E corresponding to voxels B1 to B4 are as follows. E1 = 0.0368 E2 = 0.0341 E3 = 0.3822 E4 = 0.0365 E = 0.4896

[0067] In the example in Figure 6(C), matching is performed so that the value of the evaluation function E3 corresponding to voxel B3 is high, thereby increasing the degree of matching for voxel B3. As a result, the discrepancy between the circle and star marks for voxel B3 is reduced. In this way, by appropriately setting the first weighting value, the degree of matching for voxels with a low probability of occlusion can be suitably increased, or in other words, the degree of matching for voxels with a high probability of occlusion can be suitably decreased. Similarly, for the second weighting value, by appropriately setting the second weighting value, the degree of matching for voxels with relatively high measurement accuracy can be increased, and the degree of matching for voxels with relatively low measurement accuracy can be decreased.

[0068] [Vehicle control based on standard voxels] Figure 7 is a flowchart showing the vehicle control procedure according to the reference voxel Btag. The control unit 15 repeatedly executes the process shown in the flowchart in Figure 7.

[0069] First, the vehicle position estimation unit 17 performs vehicle position estimation based on NDT matching (step S100). In this case, the vehicle position estimation unit 17 obtains the vehicle speed from the vehicle speed sensor 4 and the yaw direction angular velocity from the gyro sensor 3, and calculates the vehicle's travel distance and change in vehicle orientation based on these acquisition results. Then, the vehicle position estimation unit 17 adds the calculated travel distance and change in orientation to the estimated vehicle position one time step ago (initial value is, for example, the output value of the GPS receiver 5), and calculates the predicted position. Then, based on the calculated predicted position, the vehicle position estimation unit 17 refers to the map DB 10 and obtains voxel data of voxels present around the vehicle position. Furthermore, based on the calculated predicted position, the vehicle position estimation unit 17 divides the scan data obtained from the lidar 2 into voxels and performs NDT scan matching calculation using an evaluation function. In this case, the vehicle position estimation unit 17 calculates the evaluation function E based on equations (4) and (5). k The system then calculates an overall evaluation function E and determines the estimated parameter P that maximizes the overall evaluation function E. The vehicle position estimation unit 17 then applies the estimated parameter P to the aforementioned predicted position to calculate the estimated vehicle position at the current time.

[0070] Next, the automatic driving control unit 18 determines whether or not the position estimation accuracy has decreased (step S101). For example, the automatic driving control unit 18 determines that the position estimation accuracy has decreased if the overall evaluation function E corresponding to the estimated parameter P calculated in step S100 is smaller than a predetermined threshold. If the automatic driving control unit 18 determines that the position estimation accuracy has decreased (step S101; Yes), it proceeds to step S102. On the other hand, if the automatic driving control unit 18 determines that the position estimation accuracy has not decreased (step S101; No), it terminates the flowchart process.

[0071] In step S102, the automatic driving control unit 18 refers to the map DB 10 and obtains confidence information for voxel data corresponding to voxels around the route (step S102). In this case, for example, the automatic driving control unit 18 obtains confidence information for voxel data registered in the map DB 10 that are within a predetermined distance in the road width direction from the road on the route and within a predetermined distance from the estimated vehicle position.

[0072] Then, the automatic driving control unit 18 determines whether or not there is a voxel suitable for estimating the vehicle's position based on the confidence information acquired in step S102 (step S103). For example, the automatic driving control unit 18 determines whether or not there is a voxel in voxel data that has confidence information where the first weighting value is greater than a predetermined value and the second weighting value is less than a predetermined value. In another example, the automatic driving control unit 18 calculates a confidence index value based on an expression or table or case distinction that uses the first weighting value and the second weighting value as parameters, and determines whether or not there is a voxel in voxel data that has confidence information where the calculated confidence index value is greater than a predetermined value.

[0073] Then, if the automatic driving control unit 18 determines that there is a voxel suitable for estimating the vehicle's position (step S103; Yes), it considers the voxel suitable for estimating the vehicle's position as a reference voxel Btag and corrects the target trajectory of the vehicle so that it passes through a position suitable for measuring the reference voxel Btag (step S104). In this case, for example, the automatic driving control unit 18 corrects the target trajectory of the vehicle so that it passes through the position closest to the reference voxel Btag among positions where the reference voxel Btag is within the measurement range of the lidar 2 and where no occlusion occurs in the reference voxel Btag. In this way, the automatic driving control unit 18 can suitably improve the detection accuracy of the reference voxel Btag by the lidar 2 and suitably improve the position estimation accuracy by NDT matching by bringing the vehicle closer to the reference voxel Btag within the range in which the reference voxel Btag can be measured.

[0074] Furthermore, if there are multiple voxels suitable for estimating the vehicle's position, the automatic driving control unit 18 may, for example, use the voxel closest to the estimated vehicle position as the reference voxel Btag. In another example, the automatic driving control unit 18 may use the voxel corresponding to the confidence information with the highest confidence index value calculated based on the first weighting value and the second weighting value as the reference voxel Btag.

[0075] On the other hand, if the automatic driving control unit 18 determines that there are no voxels suitable for estimating the vehicle's position (step S103; No), it terminates the flowchart processing.

[0076] Next, a specific example based on the flowchart in Figure 7 will be explained with reference to Figure 8.

[0077] Figure 8 shows an overhead view of a vehicle on a three-lane road with a feature 40 located to the left front of the road and a feature 41 located to the right front of the road. In Figure 8, the solid arrow "L1" indicates the target trajectory before executing the flowchart in Figure 7, and the dashed arrow "L2" indicates the target trajectory after executing the flowchart in Figure 7.

[0078] In the example shown in Figure 8, after determining that the position estimation accuracy has decreased (see step S101), the automatic driving control unit 18 acquires confidence information for voxel data corresponding to voxels around the route, including voxels that overlap with features 40 and 41 (see step S102). The automatic driving control unit 18 then refers to the acquired confidence information for voxel data and determines that voxel B1, which includes a part of feature 40, is a suitable voxel for estimating the vehicle's position (see step S103). Therefore, the automatic driving control unit 18 considers voxel B1 as a reference voxel Btag and identifies a suitable position "P1" for measuring voxel B1. In this case, position P1 is, for example, the position on the road that is closest to voxel B1 among the positions on the road where voxel B1 can be measured by the lidar 2. The automatic driving control unit 18 then changes the target trajectory shown by the solid line L1 to the trajectory shown by the dashed line L2 that passes through position P1.

[0079] In this way, the automatic driving control unit 18 changes the vehicle's lane to the lane closest to the reference voxel Btag, or biases the vehicle's position within the lane towards the side closer to the reference voxel Btag while driving. As a result, the automatic driving control unit 18 can accurately measure the reference voxel Btag corresponding to highly reliable voxel data at close range using the lidar 2, thereby suitably improving the accuracy of the vehicle's position estimation.

[0080] As described above, when the in-vehicle device 1 of this embodiment determines that the position estimation accuracy has decreased, it obtains confidence information for voxel data corresponding to voxels around the route from the map DB 10 which contains voxel data. Based on the obtained confidence information, the in-vehicle device 1 sets a voxel suitable for estimating the vehicle's position as a reference voxel Btag, and outputs control information to the vehicle's electronic control unit to correct the target trajectory of the vehicle so that it passes through a position suitable for measuring the reference voxel Btag. As a result, the in-vehicle device 1 can suitably control the vehicle's driving to improve the position estimation accuracy.

[0081] [Differentiation] The following describes suitable modifications for the examples. The following modifications may be applied in combination to the examples.

[0082] (Variation 1) Instead of storing the map DB10 in the storage unit 12, the in-vehicle unit 1 may have a server device (not shown) that holds the map DB10. In this case, the in-vehicle unit 1 obtains the necessary voxel data by communicating with the server device via a communication unit (not shown).

[0083] (Modification 2) The voxel data included in map DB10 contained a first weighting value and a second weighting value as confidence information. Alternatively, the voxel data may contain only one of either the first weighting value or the second weighting value as confidence information.

[0084] In this case, the in-vehicle device 1 evaluates the evaluation function E based on either the first weighting value or the second weighting value. k The calculation and determination of the reference voxel Btag are performed. In this embodiment as well, the in-vehicle unit 1 can increase the degree of matching of highly reliable voxels, or select a highly reliable voxel as the reference voxel Btag to perform vehicle control in order to improve the accuracy of self-position estimation.

[0085] (Variation 3) The configuration of the driver assistance system shown in Figure 1 is an example, and the configuration of a driver assistance system to which the present invention can be applied is not limited to the configuration shown in Figure 1. For example, instead of having an in-vehicle unit 1, the driver assistance system may have the vehicle's electronic control unit (ECU) perform processing such as the vehicle position estimation unit 17 and the automatic driving control unit 18 of the in-vehicle unit 1. In this case, the map DB 10 may be stored in a storage unit in the vehicle, for example, and the vehicle's electronic control unit may receive update information for the map DB 10 from a server device (not shown).

[0086] (Modification 4) Voxel data is not limited to a data structure including a mean vector and a covariance matrix, as shown in Figure 3. For example, voxel data may include point cloud data measured by a map maintenance vehicle used to calculate the mean vector and covariance matrix. In this case, the point cloud data included in the voxel data is an example of "object position information" in the present invention. Furthermore, this embodiment is not limited to scan matching by NDT, and other scan matching methods such as ICP (Iterative Closest Point) may be applied. Even in this case, similar to the embodiment, the in-vehicle device 1 can suitably improve position estimation accuracy by weighting the evaluation function for each voxel that evaluates the degree of matching using a first weighting value and a second weighting value. In addition, if the position estimation accuracy decreases, the in-vehicle device 1 can suitably improve position estimation accuracy by determining a reference voxel Btag based on the first and second weighting values ​​of voxels around the route, and controlling the vehicle based on the reference voxel Btag, similar to the embodiment. [Explanation of Symbols]

[0087] 1 On-vehicle device 10 Map Database 11 Interfaces 12 Storage section 14 Input section 15 Control Unit 16. Information Output Unit

Claims

1. An acquisition unit that acquires at least precision information regarding the position information and weighting information regarding the possibility of occlusion, which are attached to voxel data in which the position information of an object for each voxel, which is a unit region, An output unit outputs control information for moving a moving object so that it approaches a voxel determined based on the accuracy information and weighting information acquired by the acquisition unit, The system includes a measuring unit for measuring the position of objects surrounding the moving body, The output device is characterized in that the weighting information is determined based on the height indicated by the position information.

2. The system further includes a position estimation unit that estimates the position of the moving object by comparing the output of a measurement unit that measures the position of surrounding objects with the position information contained in the voxel data. The output device according to claim 1, wherein the position estimation unit weights an evaluation value for evaluating the degree of matching for each voxel based on the accuracy information and the weighting information, and performs the position estimation based on the weighted evaluation value.

3. The output device according to claim 1 or 2, wherein the output unit outputs control information for moving the moving body to the lane closest to the voxel determined based on the accuracy information and the weighting information, or for moving the moving body to the side closer to the voxel within the lane in which it is traveling.

4. An acquisition unit that acquires at least precision information relating to the position information, which is attached to voxel data in which the position information of an object for each voxel, which is a unit region, An output unit outputs control information for moving a moving object so that it approaches a voxel determined based on the accuracy information acquired by the acquisition unit, The system includes a measuring unit for measuring the position of objects surrounding the moving body, The output device is characterized in that the accuracy information is determined based on the distance from the measuring vehicle that performs the measurement for generating the position information to the object to be measured and the position estimation accuracy of the measuring vehicle.

5. A control method performed by an output device, An acquisition step of acquiring at least precision information relating to the position information that is attached to voxel data in which the position information of an object for each voxel, which is a unit region, is recorded, An output step outputs control information for moving a moving body so that it approaches the voxel determined based on the accuracy information acquired in the acquisition step, The process includes a measurement step for measuring the position of objects surrounding the moving body, The accuracy information is determined by a control method based on the distance from the measuring vehicle that performs the measurement for generating the position information to the object being measured, and the position estimation accuracy of the measuring vehicle.

6. A program that is executed by a computer, An acquisition unit that acquires at least precision information relating to the position information, which is attached to voxel data in which the position information of an object for each voxel, which is a unit region, An output unit outputs control information for moving a moving object so that it approaches a voxel determined based on the accuracy information acquired by the acquisition unit, The computer is configured to function as a measuring unit for measuring the position of objects surrounding the moving object. The accuracy information is determined by a program that uses a measurement vehicle to perform measurements for generating the position information, the distance from the measurement vehicle to the object being measured, and the accuracy of the position estimation of the measurement vehicle.

7. A storage medium storing the program described in claim 6.

Images