Self-position estimation device

Abstract

To provide a self-map-generation device and a self-position-estimation device capable of estimating an error quantity of a landmark serving as an observation point and improving the quality of a self-generation point group map to actualize high-precision self-position-estimation even when a sensor has a large observation error.SOLUTION: There is provided a self-map-generation device that generates a point group map from point group data of external field information obtained by an on-vehicle sensor mounted on a vehicle itself, and the self-map-generation device comprises: a landmark detection part which detects a landmark from the point group map; a landmark information storage part which stores observation position information on the detected landmark, a landmark position correction part which corrects the position of the landmark on the basis of point group coordinates extracted from an error distribution detected by representing observation position information on the landmark in the form of a frequency distribution; and a self-map-generation part which saves, as a self-generation map, a correction map obtained from the landmark position correction part and having the position of the landmark corrected.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to Self This relates to a self-location estimation device. [Background technology]

[0002] In order to expand the scope of application of autonomous driving and driver assistance systems, it is important to acquire information from maps based on self-location estimation. However, while maps for driver assistance systems have been developed for expressways, for example, there is no prospect of developing maps for general roads or residential areas such as around one's home. Therefore, technology that can generate maps and estimate one's own location is required.

[0003] In response to this, Patent Document 1 states that "map information including the positions of fixed objects within a specified area is stored, and when recreating the map, steps (1) to (5) are repeatedly executed while moving the robot 10 within the specified area. (1) The encoder 21 is integrated to calculate the robot's own position. (2) The distance to fixed objects and movable objects within a specified viewing angle range of the robot is measured. (3) A corrected self-position is obtained from the measured distance and the distance to the fixed object obtained from the map information. (4) From the measurement values, a measurement value representing the distance to the movable object is extracted. (5) The position of the movable object is calculated from the extracted measurement value and the corrected self-position. The calculated positions of the movable objects are then aggregated to obtain new positions, and the new position obtained is output to the outside." [Prior art documents] [Patent documents]

[0004] [Patent Document 1] JP 2011-210121 A Summary of the Invention [Problem to be solved by the invention]

[0005] In the conventional technology described in Patent Document 1, the vehicle moves while automatically correcting its own position, and furthermore automatically recreates the map, thereby reducing the cost required for recreating the environmental map (point cloud map) and allowing the environmental map to be recreated in a timely manner. However, since landmark observation error information is not taken into consideration, there is a problem that the accuracy of self-location estimation is insufficient when the landmark observation error is large. When generating a map, the vehicle performs self-location estimation using landmarks registered in the map, so the landmarks used for self-location estimation need to be registered in the map with the high accuracy required for an autonomous driving / driving assistance system.

[0006] The present invention has been made in consideration of the above circumstances, and by estimating the amount of error of landmarks that serve as observation points and improving the quality of a self-generated point cloud map, it is possible to realize highly accurate self-location estimation even when the observation error of the sensor is large. Self The object of the present invention is to provide a self-location estimation device. [Means for solving the problem]

[0007] Representative of the present invention Location estimation One of the devices generates a point cloud map from point cloud data of external information obtained by on-board sensors installed in the vehicle. and estimating the vehicle's own position, which is the position of the vehicle on the point cloud map. Self to do Location estimation a landmark detection unit that detects landmarks from the point cloud map; a landmark information storage unit that stores observation position information of the detected landmarks; and a correction unit that corrects the positions of the landmarks based on point cloud coordinates extracted from an error distribution detected when the observation position information of the landmarks is frequency-distributed. and calculating the importance of the landmark based on the spread of the error distribution. a landmark position correction unit for correcting the landmark positions, and a self-generated map generation unit for storing a corrected map obtained by correcting the landmark positions obtained from the landmark position correction unit as a self-generated map. a self-location estimation unit that performs self-location estimation of the vehicle currently traveling by weighting the stored self-generated map and the corrected map newly obtained by the landmark position correction unit while the vehicle is currently traveling, by comparing the weighted map and the corrected map by the importance level; The present invention is characterized by comprising: Effect of the Invention

[0008] According to the present invention, by estimating the amount of error of landmarks that serve as observation points and improving the quality of a self-generated point cloud map, it is possible to perform highly accurate self-location estimation even when the observation error by the sensor is large.

[0009] Problems, configurations and effects other than those described above will become apparent from the following description of the embodiments. [Brief description of the drawings]

[0010] [Figure 1] 1 is a block diagram showing a configuration of a self-mapping and self-location estimation device 100 according to a first embodiment of the present invention. [Diagram 2] 1 is a functional block diagram illustrating the operation of a self-mapping and self-location estimation device 100 according to a first embodiment of the present invention. [Figure 3A] FIG. 2 is a diagram showing an outline of a method for detecting a peak position and estimating an observation error, and illustrates a currently observed point. [Figure 3B] FIG. 13 is a diagram showing an outline of a method for detecting a peak position and estimating an observation error, and is a diagram showing the form of landmark accumulated information 131. [Figure 4] 4 is a flowchart showing the operation of a landmark information storage unit 206. [Diagram 5] 11 is a flowchart showing the operation of the landmark position correction unit 207. [Figure 6] 6 is a flowchart showing the operation of a self map generating unit 208. [Figure 7] 4 is a flowchart showing the operation of a self-position estimation unit 209. [Figure 8] FIG. 11 is a block diagram showing a configuration of a self-mapping and self-location estimation device 100 equipped with a plurality of external sensors according to a second embodiment of the present invention. [Figure 9] FIG. 11 is a functional block diagram illustrating the operation of the self-mapping and self-location estimation device 100 when a plurality of external sensors are mounted, according to a second embodiment of the present invention. [Figure 10] 4 is a flowchart showing the operation of the sensor cooperative error estimator 220. [Figure 11]FIG. 11 is a block diagram showing a configuration of an automatic driving device 102 including a self-mapping and self-location estimation device 100 according to a third embodiment of the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0011] Hereinafter, an embodiment of a self-mapping device and a self-location estimation device according to the present invention will be described with reference to Figures 1 to 11. In all the figures for explaining the embodiment of the present invention, parts having the same functions are given the same reference numerals, and repeated description thereof may be omitted.

[0012] [First embodiment] 1 is a block diagram showing a configuration of a self-mapping / self-location estimation device 100 according to a first embodiment of the present invention. The self-mapping / self-location estimation device 100 is a device mounted on a vehicle, which generates a point cloud map (self-generated point cloud map) from point cloud data of external world information obtained by an on-vehicle sensor, and estimates the self-location (host vehicle position) on the point cloud map. The self-mapping / self-location estimation device 100 is composed of a group of sensors (105, 110, 111) and an on-vehicle processing device 101. The group of sensors, which are on-vehicle sensors, are connected to the on-vehicle processing device 101 by signal lines, and exchange various data with the on-vehicle processing device 101.

[0013] The in-vehicle processing device 101 includes a calculation unit 120, a RAM 130, a storage unit 140, an interface 150, and a ROM 160. The calculation unit 120 is a CPU. It may be configured to execute all or part of the calculation processing by another calculation processing device such as an FPGA. The RAM 130 is a readable and writable storage area, and operates as the main storage device of the in-vehicle processing device 101. The RAM 130 stores landmark accumulation information 131, landmark correction information 132, and a self-generated map 133, which will be described later. The ROM 160 is a read-only storage area, and stores a program, which will be described later. This program is expanded in the RAM 130 and executed by the calculation unit 120. The calculation unit 120 operates as an odometry calculation unit 121, a landmark detection unit 122, a landmark information accumulation unit 123, a landmark position correction unit 124, a self-map generation unit 125, and a self-position estimation unit 126 by reading and executing the program. The storage unit 140 is a non-volatile storage device, and operates as an auxiliary storage device for the in-vehicle processing device 101. A self-generated map 141 is stored in the storage unit 140. As will be described later, the self-generated map 141 stored in the storage unit 140 may refer to a map different from the self-generated map 133 stored in the RAM 130.

[0014] The sensor group includes an external sensor 105 that observes the surroundings of the vehicle (external information), a vehicle speed sensor 110, and a steering angle sensor 111.

[0015] The external sensor 105 is capable of observing the distance to a surrounding object (in other words, capable of functioning as a distance sensor), for example, a stereo camera. The stereo camera may be one in which two cameras are arranged side by side and separated by a certain baseline length to be one unit, or two cameras are roughly attached and can be used as a stereo camera by calibration, or two cameras are arranged vertically, or three or more cameras are used. The external sensor 105 may be a monocular camera that utilizes a motion stereo method, or may be a sensor such as LiDAR, millimeter wave radar, or sonar. In this embodiment, a stereo camera in which two cameras are arranged side by side and separated by a certain baseline length to be one unit will be described. The external sensor 105 outputs the captured image and the parallax required for distance calculation to the on-board processing device 101. Here, the external sensor 105 may output only the captured image to the on-board processing device 101, and the calculation unit 120 may calculate the parallax.

[0016] The on-board processing device 101 performs positioning of the landmark, which will be described later, using the captured image and the parallax obtained from the external sensor 105. External parameters, which are the mounting position and mounting attitude of the external sensor 105, are known and are stored in advance in the ROM 160. Alternatively, the on-board processing device 101 may estimate the position by the calculation unit 120 using the captured image, the parallax, the vehicle speed sensor 110, and the steering angle sensor 111. In addition, external parameters indicating the relative relationship between the two cameras of the stereo camera, and internal parameters such as the focal length and image sensor size of the cameras are known and are stored in advance in the ROM 160. The on-board processing device 101 can calculate the positional relationship between the external sensor 105 and the subject by using the internal parameters, external parameters, and parallax stored in the ROM 160.

[0017] The vehicle speed sensor 110 and the steering angle sensor 111 respectively measure the vehicle speed and steering angle of the vehicle in which the on-vehicle processing device 101 is mounted, and output the measured values ​​to the on-vehicle processing device 101. The on-vehicle processing device 101 uses the outputs of the vehicle speed sensor 110 and the steering angle sensor 111 to calculate the amount of movement and the direction of movement of the vehicle in which the on-vehicle processing device 101 is mounted, by a known dead reckoning technique.

[0018] (Function block configuration) 2 shows the functions of the odometry calculation unit 121, the landmark detection unit 122, the landmark information accumulation unit 123, the landmark position correction unit 124, the own map generation unit 125, and the self-location estimation unit 126 executed by the calculation unit 120 as functional blocks, and is a diagram showing the processing order of the functional blocks, as well as the flow of data between the functional blocks and the RAM 130, the ROM 160, and the storage unit 140. The own map generation / self-location estimation unit 250 as a functional block responsible for the own map generation / self-location estimation of the in-vehicle processing device 101 includes a sensor value acquisition unit 201, an odometry estimation unit 202, a landmark detection unit 203, a distance calculation unit 204, a world coordinate calculation unit 205, a landmark information accumulation unit 206, a landmark position correction unit 207, an own map generation unit 208, a self-location estimation unit 209, and a storage unit 140.

[0019] The functional block corresponding to the odometry calculation unit 121 in the calculation unit 120 is the odometry estimation unit 202, the functional block corresponding to the landmark detection unit 122 in the calculation unit 120 is the landmark detection unit 203, the functional block corresponding to the landmark information accumulation unit 123 in the calculation unit 120 is the landmark information accumulation unit 206, the functional block corresponding to the landmark position correction unit 124 in the calculation unit 120 is the landmark position correction unit 207, the functional block corresponding to the self map generation unit 125 in the calculation unit 120 is the self map generation unit 208, and the functional block corresponding to the self position estimation unit 126 in the calculation unit 120 is the self position estimation unit 209.

[0020] (Sensor value acquisition unit 201) The sensor value acquisition unit 201 acquires a signal output from the external sensor 105. In the case of a stereo camera, this is an image and a disparity. The stereo camera captures images continuously at a high frequency, for example, 30 times per second. The images and disparity acquired by the stereo camera are transmitted to the in-vehicle processing device 101 every time an image is captured. The sensor value acquisition unit 201 receives this at a constant frequency, and outputs the received images and disparity to the landmark detection unit 203. The subsequent processing units 203 to 209 operate every time they receive the image and disparity.

[0021] (Odometry estimation unit 202) The odometry estimation unit 202 estimates vehicle motion using the vehicle speed and steering angle transmitted from the vehicle speed sensor 110 and steering angle sensor 111. For example, a known dead reckoning technique may be used, a known visual odometry technique using a camera may be used, or a known Kalman filter or the like may be used in combination. If a GPS is provided, such as in the autonomous driving system shown in FIG. 11 described later, information from the GPS may also be utilized.

[0022] (Landmark detection unit 203) The landmark detection unit 203 detects landmarks using the image and parallax obtained from the sensor value acquisition unit 201. A landmark is an object having a feature identifiable by a sensor, such as a lane mark, which is a type of road paint, a stop line, a crosswalk, other regulatory signs, and a building wall, which is an obstacle that impedes the travel of a vehicle. In this embodiment, vehicles and humans, which are moving bodies, are not included in the landmarks. Based on information input from the external sensor 105 via the sensor value acquisition unit 201 (more specifically, a point cloud map generated from point cloud data of the external information obtained by the external sensor 105), landmarks existing around the vehicle, that is, objects having a feature identifiable by the sensor, are detected. The landmark information may be obtained in pixel units or may be obtained by grouping as an object. The landmark may or may not be identified as, for example, a lane mark or a crosswalk by image recognition. The landmark information obtained here is output to the distance calculation unit 204.

[0023] (Distance calculation unit 204) The distance calculation unit 204 calculates the distance to the landmark based on the landmark information obtained by the landmark detection unit 203. The distance is detected for each pixel where the landmark exists or for grouped objects. In the case of a stereo camera, the distance can be measured by the principle of triangulation from the parallax value, the internal parameters stored in the ROM 160, and the external parameters between the cameras. Furthermore, the three-dimensional coordinates of the object reflected in the corresponding pixel can be measured from the distance information. Using this, the corresponding three-dimensional coordinates (observation position information) are measured for the pixel where the landmark exists or for the grouped object. However, the parallax is calculated by processing such as block matching, and always contains an error. This error causes an error in the distance, and the three-dimensional coordinates of the object contain an observation error. The observation error of this three-dimensional coordinate has the tendency to occur in the direction of a straight line connecting the origin and the object in general cameras including stereo cameras. When other external sensors (distance sensors) are used, a similar error may occur. When a self-map (self-generated map) is generated by a self-map generating unit 208 (described later) and a self-location estimating unit 209 (described later) estimates the self-location in a state where the error is large, the accuracy of the self-location estimation decreases due to the influence of the error. Therefore, it is desirable to reduce the error by some method. Moreover, the three-dimensional coordinates are observed values ​​in the sensor coordinate system and are relative coordinate values ​​from the sensor. The relative coordinate values ​​obtained here are output to a world coordinate calculating unit 205.

[0024] (World coordinate calculation unit 205) The world coordinate calculation unit 205 can convert the distance information (observation position information) of the landmark, which is a relative coordinate value from the sensor obtained by the distance calculation unit 204, into a world coordinate value using the motion information of the vehicle obtained by the odometry estimation unit 202 and external parameters stored in the ROM 160 that indicate the mounting position and mounting attitude of the camera. When the vehicle is not equipped with a GPS and estimation is performed using dead reckoning or visual odometry, the world coordinate value is a world coordinate value based on a certain coordinate or axis. For example, the position where the in-vehicle processing device 101 is started may be taken as the origin, the direction in which the vehicle proceeded immediately afterwards as the X axis, and the axes perpendicular thereto as the Y axis and the Z axis. When the vehicle is equipped with a GPS, as in the case of an automatic driving system shown in FIG. 11 described later, absolute world coordinate values ​​such as latitude and longitude can be obtained. The world coordinate values ​​of each landmark obtained here are output to the landmark information accumulation unit 206.

[0025] (Landmark information storage unit 206) The landmark information storage unit 206 accumulates the measurement (observation) position information of each landmark obtained by the world coordinate calculation unit 205 in a chronological order. This information is utilized by the landmark position correction unit 207 described later to reduce the position measurement error of each landmark. The landmark information storage unit 206, for example, votes the measurement position information (world coordinate value) of each landmark obtained by the world coordinate calculation unit 205 in a chronological order on a map called a grid map, which is a voxel-shaped partition of space. The voting here means a process of incrementing the value of a voxel by +1 when one observation value contained in the corresponding voxel is obtained. When this is repeated every time the sensor value acquisition unit 201 receives a sensor value (signal) (i.e., when the frequency distribution is performed), a difference (in the frequency distribution) appears between a location with a high number of observations and a location with a low number of observations. In the case of a stereo camera, the observation value has a measurement distance error resulting from a parallax error, and this is reflected in the voting value of the grid map, and the error distribution for each observed object (i.e., the error distribution detected when the measurement position information is frequency distributed) appears on the grid map. This will be described later with reference to FIG. 3A and FIG. 3B. In this process, landmark information is accumulated in a time series manner, like voting on a grid map. The grid map may be two-dimensional or three-dimensional. In order to accumulate the landmark information in a time series manner, when voting of the world coordinate values ​​for all the landmarks obtained by the sensor value acquisition unit 201 at a certain time is completed, the information accumulated in the RAM 130 is stored as the landmark accumulation information 131 (FIG. 1). At the next time, when the sensor value acquisition unit 201 receives a sensor value (signal), the landmark accumulation information 131, which is the accumulated landmark information, is read from the RAM 130, and the observation value (world coordinate value) at the current time is accumulated in the read information. The processing flow of the landmark information accumulation unit 206 will be described later with reference to FIG. 4. Next, the process proceeds to the landmark position correction unit 207.

[0026] (Landmark position correction unit 207) The landmark position correction unit 207 corrects the position error of the landmark detection by using the landmark accumulation information 131 (FIG. 1) stored in the RAM 130. As described above, when the landmark information accumulation unit 206 accumulates the landmark information, the error distribution for each observed object (i.e., the error distribution detected when the measurement position information is frequency-distributed) can be visualized. In addition, this error distribution has a tendency for the error to spread in the direction of the straight line connecting the sensor and the observed object (the line of sight or the light direction) due to the characteristics of the sensor (due to the parallax error of the distance sensor). Therefore, the landmark accumulation information 131 (FIG. 1) is scanned in the direction of the straight line connecting the sensor and the observed object (the line of sight or the light direction), and the peak position of the error distribution is detected as the landmark point cloud coordinates, thereby estimating the true observed position of the object. This will be explained further later with reference to FIG. 3A and FIG. 3B. The estimated observed position (= peak position) is set as the landmark correction position. At the same time, the larger the magnitude of the error distribution, the lower the observation reliability (also called importance) for the observed object. Therefore, the landmark position correction unit 207 acquires the spread of the error distribution (corresponding to the estimated error amount) and utilizes it as the reliability (importance) of the target observation object in subsequent processing. The landmark corrected position and the magnitude of the spread of the error distribution of the corresponding landmark (estimated error amount) are stored in the RAM 130 as the landmark correction information 132 (FIG. 1). The processing flow of the landmark position correction unit 207 will be explained later with reference to FIG. 5. Next, the process proceeds to the self map generation unit 208.

[0027] (Self-map generation unit 208) The self map generating unit 208 generates a map (more specifically, a corrected map in which the positions of the landmarks are corrected) using the landmark accumulation information 131 and the landmark correction information 132 obtained by the landmark position correction unit 207. The map may be expressed as a point cloud corresponding to the landmark points, or may be expressed in the form of recognized objects and their coordinates. When outputting as a point cloud, the self map outputs to the self-generated map, based on the peak position information obtained by the landmark correction information 132, landmarks in a certain range centered on the peak position in the landmark accumulation information 131 as a landmark point cloud.

[0028] The operation of the self map generation unit 208 changes depending on the mode of the self map generation / self position estimation unit 250. There are two modes: a self map generation mode and a self position estimation mode. The two modes are assumed to be determined in some way when the self map generation / self position estimation unit 250 is operated. For example, mode information may be input by some means from the interface 150 (FIG. 1).

[0029] (Self-map generation mode) The self map generation mode is intended for self map generation. In the self map generation mode, the subsequent self position estimation unit 209 is not executed. When the self map generation unit 208 receives a map save command, it stores (saves) the entire range obtained from the landmark accumulation information 131 in the memory unit 140 as the self-generated map 141 (FIG. 1). After storage, the process does not proceed to the subsequent self position estimation unit 209, and ends.

[0030] (Self-location estimation mode) The self-location estimation mode is intended to read out the self-generated map 141 (FIG. 1) previously stored in the storage unit 140 by the self-map generation mode, and estimate the self-location on the self-generated map 141. In the self-location estimation unit 209 at the latter stage, the self-generated map previously stored in the storage unit 140 is compared with the self-generated map (=corrected map) outputted (from the self-map generation unit 208) in the self-location estimation mode from the current observation value. In the self-location estimation mode, the self-generated map (=corrected map) outputted (from the self-map generation unit 208) is a partial map of a range going back a predetermined distance from the current vehicle position. By using a partial map, there are effects of preventing matching errors, shortening the matching time, and reducing the influence of odometry errors. The partial map is stored (saved) in the RAM 130 as the self-generated map 133 (FIG. 1), and is outputted to the self-location estimation unit 209. Next, the process proceeds to the self-location estimation unit 209. The process flow of the self-map generation unit 208 will be described later with reference to FIG. 6.

[0031] (Self-position estimation unit 209) The self-position estimation unit 209 operates only in the self-position estimation mode. The self-position estimation unit 209 estimates the self-position on the self-generated map 141 by comparing the self-generated map (partial map, corrected map) 133 obtained from the self-map generation unit 208 with the self-generated map 141 generated in the past and stored in the storage unit 140. In the case of a point cloud, for example, the ICP (Iterative Closest Point) algorithm, which is a known point cloud matching technique, can be used for the comparison. This makes it possible to obtain a coordinate transformation amount for transforming the self-generated map (partial map, corrected map) 133 obtained from the self-map generation unit 208 onto the self-generated map 141, and the self-position on the self-generated map 141 can be obtained from the current vehicle position that has been coordinate-transformed based on the coordinate transformation amount. The estimated self-position is output, and the processing of the self-map generation / self-position estimation unit 250 is completed. The processing flow of the self-position estimation unit 209 will be described later with reference to FIG. 7.

[0032] FIG. 3A is a diagram showing a currently observed point. If the external sensor 105 is a camera, the image captured by the camera will be similar to this figure. Electric poles 310, 311, and 312 exist at this point. FIG. 3B is a diagram showing the state of the landmark accumulated information 131 generated by the landmark information accumulation unit 206 at this point. In this example, for ease of understanding, the grid map is shown as a two-dimensional case in which the height direction is compressed. Therefore, the figure is viewed from a bird's-eye view. 320, 321, and 322 in FIG. 3B represent the contour lines of the voting values ​​of the landmark accumulated information 131, and correspond to the observation results of the electric poles 310, 311, and 312, respectively. 330 in FIG. 3B represents the sensor position.

[0033] If the observation error of the utility poles 310, 311, and 312 is 0, the observation results would normally only appear in a range equivalent to the diameter of the utility poles on a two-dimensional grid map. However, if there is an observation error, due to the error characteristics of a sensor such as a stereo camera (due to the parallax error of a distance sensor), the error will be distributed in the direction of a straight line (the direction of line of sight or the direction of light) connecting the sensor position 330 and the utility poles 310, 311, and 312, and the contour lines of the voting values ​​of the landmark accumulation information 131 will show an error distribution such as 320, 321, and 322. The error due to the sensor basically has 0 as the mode, and the width and shape of the distribution change depending on the sensor, object, weather, and sunlight conditions, and the error characteristics are reflected in the peak position and shape of each distribution, as shown in Figure 3B. Auxiliary lines 340, 341, and 342 are straight lines connecting the sensor position 330 and the observation result corresponding to utility pole 310, auxiliary lines 343, 344, and 345 are straight lines connecting the sensor position 330 and the observation result corresponding to utility pole 311, and auxiliary lines 346, 347, and 348 are straight lines connecting the sensor position 330 and the observation result corresponding to utility pole 312. By scanning each of these straight lines and detecting the peak position of the voting value of the landmark accumulation information 131, the true position of the landmark can be estimated even if the sensor, object, weather, and sunlight conditions change. Each of these auxiliary lines is merely a representative line for the purpose of explanation, and scanning and detection with finer resolution may be performed.

[0034] Moreover, the larger the distribution, the larger the error, meaning that the observed value is less reliable (low observation reliability). Therefore, by determining the magnitude of this distribution on each line, the reliability (importance) of each landmark can be calculated. In this example, the distribution 321 for utility pole 311 has the largest error (estimated error amount) and is therefore less reliable (low observation reliability), whereas the distribution 311 for utility pole 310 has the smallest error (estimated error amount) and is therefore more reliable (high observation reliability).

[0035] Peak detection can be achieved by smoothing the voting values ​​of the landmark accumulation information 131 on the scanning line and then using a known zero crossing method or the like. The magnitude of the distribution can be obtained by determining the width from the peak position where the voting value of the landmark accumulation information 131 is equal to or greater than a predetermined value. In this example, the case where the grid map is two-dimensional is shown, but the same applies to the case where it is three-dimensional.

[0036] (Operation of landmark information storage unit 206) 4 shows a flowchart showing the operation of the landmark information accumulation unit 206. The landmark information accumulation unit 206 executes the following operations every time it receives an execution command from the world coordinate calculation unit 205. The execution subject of each step described below is the calculation unit 120.

[0037] In step S401, the landmark accumulated information acquisition unit acquires the landmark accumulated information 131 obtained up to the previous time and stored in the RAM 130. Next, the process proceeds to step S402.

[0038] The world coordinate acquisition unit in step S402 acquires the world coordinate value of the landmark at that time calculated by the world coordinate calculation unit 205. Next, the process proceeds to step S403.

[0039] In step S403, the landmark information adding unit updates the landmark accumulation information 131 using the landmark accumulation information 131 obtained in step S401 and the world coordinate value of the landmark at the time calculated by the world coordinate calculation unit 205 obtained in step S402. As described in the explanation of Figs. 3A and 3B, the landmark accumulation information 131 is a space divided into a grid map or the like, to which landmark points are voted. The world coordinate value of the landmark at the time calculated by the world coordinate calculation unit 205 is additionally voted and updated in this space. This makes it possible to reflect the landmark information observed at the time in the landmark accumulation information 131. The updated landmark accumulation information 131 is stored in the RAM 130 for use by the landmark information accumulation unit 206 and the landmark position correction unit 207 at the next time onwards, replacing the old one, and the flowchart of Fig. 4 ends.

[0040] (Operation of landmark position correction unit 207) 5 shows a flowchart showing the operation of the landmark position correction unit 207. The landmark position correction unit 207 executes the following operations every time it receives an execution command from the landmark information storage unit 206. The calculation unit 120 is responsible for executing each step described below.

[0041] In step S501, the landmark accumulated information acquisition unit acquires the landmark accumulated information 131 calculated in the landmark information accumulation unit 206 and stored in the RAM 130. Next, the process proceeds to step S502.

[0042] In step S502, the landmark accumulation completion determination unit determines the landmark grids for which accumulation has been completed. In world coordinate values, a location that has gone outside the measurement range of the current position of the sensor is determined as a location for which landmark accumulation has been completed. The grids for which landmark accumulation has been completed are the targets of subsequent processing. Next, proceed to step S503.

[0043] In the peak position scanning unit in step S503, as described with reference to Fig. 3A and Fig. 3B, a scanning line is set on a straight line connecting each landmark and the sensor position on the landmark accumulation information 131 acquired in step S501. The scanning line is set for each grid for which it is determined that the accumulation of the landmarks obtained in step S502 has been completed. Next, proceed to step S504.

[0044] In step S504, the landmark output position calculation unit scans the landmark accumulation information 131 acquired in step S501 along the scanning line set in step S503, and detects the peak position of the error distribution as point cloud coordinates. Peak detection can be achieved by smoothing the voting values ​​of the landmark accumulation information 131 on the scanning line and then using a known zero crossing method or the like. Once the peak position is calculated, it is stored in the RAM 130 as the landmark correction information 132. Multiple peaks may be obtained from one scanning line. Next, proceed to step S505.

[0045] Here, the peak position in this specification may be a pure vertex position obtained by performing some kind of signal processing such as smoothing on the voting values ​​of the landmark accumulation information 131 on the scanning line, or may be all the remaining coordinates obtained by performing some kind of signal processing such as smoothing on the voting values ​​of the landmark accumulation information 131 on the scanning line and then threshold processing, or may be the average value of the remaining coordinates. The signal processing method for peak detection and the number of points output may be any method that limits the remaining points to those near the peak position.

[0046] The landmark distribution calculation unit in step S505 scans the scanning line set by the peak position scanning unit in step S503, and calculates the magnitude of error distribution for each peak position obtained by the landmark output position calculation unit in step S504. The magnitude of error distribution is obtained by, for example, determining the width from the peak position where the voting value of the landmark accumulation information 131 is equal to or greater than a predetermined value. The obtained magnitude of error distribution for each peak position is stored in RAM 130 as landmark correction information 132, and the flowchart in FIG. 5 ends.

[0047] (Operation of the self map generating unit 208) 6 shows a flowchart showing the operation of the own map generating unit 208. The own map generating unit 208 executes the following operations every time it receives an execution command from the landmark position correcting unit 207. The execution subject of each step described below is the calculation unit 120.

[0048] In step S601, the landmark information acquisition unit acquires the landmark accumulation information 131 and the landmark correction information 132 stored in the RAM 130. Next, the process proceeds to step S602.

[0049] In the range setting unit in step S602, the operation is switched depending on the mode of the self-mapping generation / self-location estimation unit 250. In the self-mapping generation mode, the target range is the entire map generated by the landmark accumulation information 131. In the self-location estimation mode, the target range is a range going back a predetermined distance from the current vehicle position. Therefore, a portion (partial map) generated by the landmark accumulation information 131 is the target range. This has the effect of preventing matching errors in the subsequent self-location estimation unit 209, shortening the matching time, and reducing the influence of odometry errors. Once the target range has been determined, proceed to step S603.

[0050] The map output unit in step S603 uses the landmark accumulation information 131 and landmark correction information 132 obtained by the landmark information acquisition unit in step S601, and the target range obtained by the range setting unit in step S602, and outputs a certain range of landmarks centered on the peak position in the landmark accumulation information 131 as a landmark point cloud to the self-generated map (141 or 133) based on the peak position information obtained in the landmark correction information 132. For example, the landmark accumulation information 131 is scanned in the scanning line direction from the peak value, and a range having a value of 80% of the peak value is output as the landmark point cloud coordinates. This 80% is just an example, and various values ​​may be set in practice. Alternatively, the range may be narrowed and only one peak value may be output as the landmark point cloud coordinates.

[0051] In the case of the self-map generation mode, if a map recording end command is received, the obtained landmark point cloud coordinates are stored in the memory unit 140 as the self-generated map 141, and the flow of Fig. 6 is terminated. Furthermore, the self-map generation / self-location estimation unit 250 also terminates. If a map recording end command is not received, the obtained landmark point cloud coordinates are stored in the RAM 130 as the self-generated map 133, and the flow of Fig. 6 is terminated. The process does not proceed to the next self-location estimation unit 209, but returns to the sensor value acquisition unit 201. In the case of the self-location estimation mode, the obtained landmark point cloud coordinates are stored in the RAM 130 as the self-generated map 133, and the flow of Fig. 6 is terminated.

[0052] (Operation of self-position estimation unit 209) 7 is a flowchart showing the operation of the self-location estimation unit 209. The self-location estimation unit 209 executes the following operation every time it receives an execution command from the self-map generation unit 208. The execution subject of each step described below is the calculation unit 120. This process is executed only when the self-map generation / self-location estimation unit 250 is in the self-location estimation mode.

[0053] In step S701, the past map information acquisition unit acquires the self-generated map 141 stored in the storage unit 140. The self-generated map 141 is a map acquired by past travel. Once acquired, the process proceeds to the next step S702.

[0054] In step S702, the current map information acquisition unit acquires the self-generated map (partial map, corrected map) 133 observed during the current travel, which is stored in the RAM 130 by the self-map generation unit 208. After acquisition, the process proceeds to the next step S703.

[0055] In step S703, the weight calculation unit acquires the landmark correction information 132 stored in the RAM 130, and acquires the error distribution magnitude information for each peak position of the landmark correction information 132. Based on this error distribution magnitude information (corresponding to the estimated error amount), weight information (importance or reliability) of each landmark when estimating the self-position is calculated. As a setting policy, the larger the distribution, the smaller the landmark weight is set. The calculation formula for the weight for the distribution magnitude may be freely set while adhering to the policy. Next, proceed to step S704.

[0056] In the map matching section in step S704, the past self-generated map 141 acquired in the past map information acquisition section in step S701 is matched with the landmarks of the self-generated map (partial map, corrected map) 133 observed during the current drive acquired in the current map information acquisition section in step S702, thereby calculating the self-position on the past self-generated map 141 during the current drive. In the case of a point cloud, for example, the ICP (Iterative Closest Point) algorithm, which is a known point cloud matching technique, can be used for the matching. When the ICP algorithm is used, the self-position is calculated by reflecting the weight (importance or reliability) of each landmark obtained in the load calculation section in step S703. In the ICP algorithm, the point clouds are matched by repeating the process of finding the correspondence between each point and minimizing the distance error in the correspondence, and at this time, the weight is reflected in the distance error function for minimization. The error of a point with a large weight is calculated as a relatively large value, and a force that tries to minimize it is applied, so that points with a large weight are more likely to reduce the error than other points. This makes it possible to reduce the error for more reliable points with larger weights. By comparing the maps, it is possible to determine the amount of coordinate transformation for transforming the self-generated map (partial map, corrected map) 133 obtained from the self-map generating unit 208 onto the self-generated map 141, and it is possible to obtain the self-position on the self-generated map 141 from the current vehicle position that has been coordinate-transformed based on the amount of coordinate transformation. The self-position on the self-generated map 141 is output, and the flow of Fig. 7 is terminated. After termination, the flow returns to the sensor value acquiring unit 201.

[0057] (Action and effect) According to the above-described first embodiment, the following advantageous effects can be obtained. That is, the self-mapping / self-location estimation device 100 of the first embodiment described above generates a point cloud map from point cloud data of external environment information obtained by an on-board sensor mounted on the host vehicle, and includes: a landmark detection unit 203 which detects landmarks from the point cloud map; a landmark information storage unit 206 which stores observation position information of the detected landmarks; a landmark position correction unit 207 which corrects the positions of the landmarks based on point cloud coordinates extracted from an error distribution detected when the observation position information of the landmarks is frequency-distributed (more specifically, a peak position of the error distribution detected when the observation position information of the landmarks is frequency-distributed is extracted as a point cloud coordinate, and based on the peak position); and a self-map generating unit 208 which stores (in the storage unit 140 or RAM 130) the corrected map obtained by correcting the positions of the landmarks obtained from the landmark position correction unit 207 as a self-generated map.

[0058] The vehicle-mounted sensor is a distance sensor (e.g., a stereo camera) composed of multiple cameras, and the error distribution occurs in the line of sight (light direction) due to parallax error of the distance sensor. The landmark position correction unit 207 scans the distance sensor in the line of sight (light direction) and extracts the peak position of the error distribution as point cloud coordinates.

[0059] The self-mapping / self-location estimation device 100 also includes a self-location estimation unit 209 that performs self-location estimation of the vehicle currently traveling by comparing the stored self-generated map with the corrected map newly obtained from the landmark position correction unit 207 while the vehicle is currently traveling.

[0060] In addition, the self-location estimation unit 209 calculates the importance (reliability) of the landmark using the estimated error amount (information on the magnitude of error distribution for each peak position) obtained from at least one of the corrected map obtained from the landmark position correction unit 207 or the self-generated map (S703: load calculation unit), and performs self-location estimation reflecting the obtained importance (reliability).

[0061] Therefore, by estimating the amount of error in the landmarks that serve as observation points and improving the quality of the self-generated point cloud map, it is possible to achieve highly accurate self-location estimation even when the observation error from the sensor is large.

[0062] [Second embodiment] 8 is a block diagram showing a configuration of a self-mapping / self-location estimation apparatus 100 according to a second embodiment of the present invention. The self-mapping / self-location estimation apparatus 100 is made up of a group of sensors (106, 107, 110, 111) and an on-vehicle processing device 101. The group of sensors, which are on-vehicle sensors, are connected to the on-vehicle processing device 101 by signal lines, and exchange various types of data with the on-vehicle processing device 101.

[0063] The on-vehicle processing device 101 includes a calculation unit 120, a RAM 130, a storage unit 140, an interface 150, and a ROM 160. The calculation unit 120 differs from the first embodiment in FIG. 1 in that it has a sensor cooperative error estimation unit 127. Also, the second embodiment differs from the first embodiment in FIG. 1 in that it simultaneously has, as a sensor group, a stereo camera 106 and a monocular camera (hereinafter, sometimes simply referred to as a camera) 107, which are two different types of external sensors that observe the surroundings of the vehicle (external environment information). In this embodiment, the stereo camera 106 and the monocular camera 107 are used, but a combination of two other different types of sensors may be used.

[0064] (Function block configuration) FIG. 9 is a functional block diagram showing the operation of the self-mapping / self-location estimation device 100 when multiple external sensors are mounted. The functions of the odometry calculation unit 121, the landmark detection unit 122, the landmark information storage unit 123, the sensor cooperative error estimation unit 127, and the self-mapping / self-location estimation unit 250 executed by the calculation unit 120 are shown as functional blocks, and the processing order of the functional blocks and the data flow between the functional blocks and the RAM 130, the ROM 160, and the storage unit 140 are shown. The odometry calculation unit 121, the landmark detection unit 122, the landmark information storage unit 123, and the self-mapping / self-location estimation unit 250 are the same as those shown in the first embodiment of FIG. 1, so the description will be omitted. However, the external sensor of the self-mapping / self-location estimation unit 250 is the stereo camera 106. The sensor value acquisition unit 201 in FIG. 9 acquires information from the camera 107.

[0065] (Sensor Cooperative Error Estimation Unit 220) The sensor cooperative error estimation unit 220 estimates the error amount (landmark detection error amount) of the landmark accumulated information 131 obtained from the camera 107, using the landmark accumulated information 131 by the camera 107 obtained from the landmark information accumulation unit 206 and the landmark correction information 132 by the stereo camera 106 obtained from the self map generation / self position estimation unit 250 (corresponding to the corrected map created based on the stereo camera 106). For example, in the case of the motion stereo method of the monocular camera 107, it is sometimes difficult to obtain accurate peak positions and the magnitude of error distribution because the landmark information cannot be obtained densely in practice and the votes for the landmark accumulated information 131 are not collected. On the other hand, the stereo camera 106 can obtain landmark information densely and has good accuracy, so the landmark correction information 132 obtained from the stereo camera 106, which is a sensor with high accuracy, is utilized to estimate the error amount of the landmark accumulated information 131 obtained from the camera 107, which is a sensor with low accuracy. If the amount of error of the landmark accumulation information 131 can be estimated, the importance (reliability) of the landmark can be calculated, and as described above, this can be reflected as the weight (importance or reliability) of each landmark during self-location estimation, improving the accuracy of self-location estimation. However, this process is premised on the fact that the sensors are able to observe a common object. For the common observation object, the landmark corrected position obtained from the landmark correction information 132 estimated by the sensor with higher accuracy is regarded as the true value, and the deviation of the distance from the true value of the observation position of the observation value of the common observation object acquired by the sensor with lower accuracy is regarded as the amount of error in the sensor with lower accuracy.

[0066] (Operation of the sensor cooperative error estimator 220) 10 is a flowchart showing the operation of the sensor coordination error estimation unit 220. The sensor coordination error estimation unit 220 executes the following operations every time it receives an execution command from the landmark information accumulation unit 206. The execution subject of each step described below is the calculation unit 120.

[0067] 4, the landmark accumulated information acquisition unit in step S401 acquires the landmark accumulated information 131 obtained from the camera 107 and stored in the RAM 130. Next, the process proceeds to step S1002.

[0068] A landmark correction information acquisition unit in step S1002 acquires landmark correction information 132 (corrected map) obtained from the stereo camera 106 and stored in the RAM 130. Next, the process proceeds to step S1003.

[0069] In step S1003, a common object determination unit determines an object commonly observed by the stereo camera 106 and the camera 107. This is based on the premise that the sensors have a common field of view. The determination is made using a predetermined known image recognition method. When the determination of all common objects in the observation range is completed, the process proceeds to step S1004.

[0070] In step S1004, the landmark output position collaborative calculation unit corrects the landmark position of camera 107, which is the sensor with lower accuracy, using landmark accumulation information 131 obtained from camera 107, landmark correction information 132 (corrected map) obtained from stereo camera 106, and the common object determination result obtained by the common object determination unit in step S1003. In landmark correction information 132 obtained by stereo camera 106, which is the sensor with higher accuracy, the peak position of the landmark determined to be a common object is set as the true object detection coordinate. The peak position of landmark accumulation information 131 obtained from camera 107 is ignored. Next, proceed to step S1005.

[0071] In step S1005, the landmark distribution collaborative calculation unit estimates the error distribution of the landmarks using the landmark correction information 132 obtained in step S1004 and the landmark accumulation information 131 obtained from the camera 107. For grids in the landmark accumulation information 131 that are within a predetermined distance from each landmark correction position and have a voting value equal to or greater than a predetermined value, the errors from the landmark correction position are measured and distributed as a frequency. The width of this frequency distribution is regarded as the estimated error distribution of the landmark (corresponding to the landmark detection error amount).

[0072] (Action and effect) According to the second embodiment, the following advantageous effects can be obtained. That is, the self-mapping / self-location estimation device 100 of the second embodiment described above further includes a sensor cooperative error estimation unit 220 that uses the corrected map created based on the on-vehicle sensor (e.g., the stereo camera 106) to estimate the amount of detection error of the landmark (observation position information of the landmark) detected by a sensor other than the on-vehicle sensor (e.g., the monocular camera 107).

[0073] Therefore, when there are multiple external sensors that have a common observation range, the sensor with higher accuracy can be used to correct the landmark positions of the sensor with lower accuracy and estimate the error distribution (detection error amount) of the landmarks, thereby improving the accuracy of the self-generated map and increasing the accuracy of self-location estimation by using the error distribution (detection error amount).

[0074] [Third embodiment] 11 is a block diagram showing a configuration of an automatic driving device 102 including the self-mapping / self-location estimation device 100 according to the third embodiment of the present invention. The automatic driving device 102 includes a modified configuration of the self-mapping / self-location estimation device 100, a communication device 113, a display device 114, and a group of vehicle control devices (170-173) that control the vehicle.

[0075] The self-mapping / self-location estimation device 100 is composed of a group of sensors (105, 110, 111, 112) and an on-board processing device 101. The group of sensors which are on-board sensors, the group of vehicle control devices, the communication device 113, and the display device 114 are connected to the on-board processing device 101 via signal lines, and exchange various data with the on-board processing device 101. The self-mapping / self-location estimation device 100 differs from the first embodiment in FIG. 1 only in that it is equipped with a GPS receiver 112.

[0076] The GPS receiver 112 receives signals from multiple satellites that make up a satellite navigation system, and calculates the position of the GPS receiver 112, i.e., the latitude and longitude, by calculation based on the received signals. The accuracy of the latitude and longitude calculated by the GPS receiver 112 does not need to be high, and may include an error of, for example, several meters to 10 meters. The GPS receiver 112 outputs the calculated latitude and longitude to the on-board processing device 101.

[0077] The communication device 113 is used to wirelessly transmit and receive information between devices outside the vehicle and the in-vehicle processing device 101. For example, when the user is outside the vehicle, the communication device 113 communicates with a mobile terminal carried by the user to transmit and receive information. The communication device 113 is not limited to communicating with the user's mobile terminal.

[0078] The display device 114 is, for example, a liquid crystal display, and displays information output from the on-vehicle processing device 101 .

[0079] The vehicle control device 170 controls the steering device 171, the driving device 172, and the braking device 173 based on information output from the on-board processing device 101, for example, the current (driving) self-position on the self-generated map 141 output from the self-position estimation unit 126. The steering device 171 operates the steering of the vehicle. The driving device 172 applies a driving force to the vehicle. The driving device 172 increases the driving force of the vehicle, for example, by increasing the target rotation speed of an engine equipped in the vehicle. The braking device 173 applies a braking force to the vehicle.

[0080] In the automatic driving device 102, the self-generated map 141 stored in the memory unit 140 may be stored (saved) in a server or the like via the communication device 113 so that it can be used by other vehicles. Also, the self-generated map 141 stored in a server via the communication device 113 by the automatic driving device 102 mounted on another vehicle may be acquired via the communication device 113, stored in the memory unit 140, and used for self-position estimation as a self-generated map of the vehicle itself.

[0081] In the automatic driving device 102, the reception position (estimated absolute coordinates) of the GPS receiver 112 at each time is attached to a self-generated map, and when estimating the self-position, a map with a value close to the current GPS value is searched for and a map to be checked is selected. Even if the GPS cannot be received, the odometry calculation unit 121 refers to the GPS to estimate the absolute position (absolute coordinates), and the absolute position (absolute coordinates) is estimated from the most recently observed GPS value.

[0082] The automatic driving device 102 (the vehicle control device 170) estimates the self-generated map 141 generated by the self-mapping and self-location estimation device 100 and its own position on the self-generated map 141 by a self-location estimation unit 126 provided in the self-mapping and self-location estimation device 100, and automatically drives the vehicle according to the estimated self-location.

[0083] (Action and effect) According to the third embodiment, the following advantageous effects can be obtained. That is, the self-mapping and self-location estimation device 100 of the above-mentioned third embodiment includes a vehicle control device 170 (automatic driving device 102) that automatically drives the vehicle according to the self-generated map generated by the self-mapping and self-location estimation device 100 and the self-location estimated by the self-location estimation unit 126 (209) included in the self-mapping and self-location estimation device 100 based on the self-generated map.

[0084] In addition, in order to enable other vehicles to utilize the self-generated map 141, the self-mapping / self-location estimation device 100 stores the self-generated map 141 in a server via the communication device 113, or obtains the self-generated map 141 generated by another vehicle via the communication device 113 and can utilize it as the self-generated map 141 of the vehicle itself.

[0085] Moreover, the self-mapping / self-location estimation device 100 assigns absolute coordinates estimated from GPS and odometry to the self-generated map.

[0086] According to the third embodiment, highly accurate automatic driving of the vehicle is possible based on the highly accurate estimated self-location. In addition, by sharing the self-generated map with other vehicles via a server, highly accurate self-location estimation is possible even when there is little information (driving experience).

[0087] The present invention is not limited to the above-mentioned embodiment, and various modifications are included. For example, the above-mentioned embodiment has been described in detail to explain the present invention in an easy-to-understand manner, and is not necessarily limited to those having all the configurations described. Other aspects conceivable within the scope of the technical idea of ​​the present invention are also included in the scope of the present invention. In addition, it is possible to replace a part of the configuration of one embodiment with the configuration of another embodiment, and it is also possible to add the configuration of another embodiment to the configuration of one embodiment. In addition, it is possible to add, delete, or replace other configurations with respect to a part of the configuration of each embodiment. In addition, the self-mapping and self-location estimation device 100 may be provided with an input / output interface (not shown), and a program may be read from another device via the input / output interface and a medium available to the self-mapping and self-location estimation device 100 when necessary. Here, the medium refers to, for example, a storage medium that is detachable from the input / output interface, or a communication medium, i.e., a network such as a wired, wireless, or optical network, or a carrier wave or digital signal that propagates through the network. In addition, some or all of the functions realized by the program may be realized by a hardware circuit or an FPGA. In addition, each of the above-mentioned configurations, functions, etc. may be realized by software by a processor interpreting and executing a program that realizes each function. Information such as programs, tables, and files that realize each function can be stored in a memory, a recording device such as a hard disk or SSD (Solid State Drive), or a recording medium such as an IC card, SD card, or DVD. [Explanation of symbols]

[0088] 100...Self-mapping and self-location estimation device 101...In-vehicle processing device 102...Automatic driving device 105...External sensor 106…Stereo camera 107...Camera 110…Vehicle speed sensor 111...Steering angle sensor 112…GPS receiver 113...Communication equipment 114...Display device 120...Arithmetic section 121...Odometry calculation unit 122…Landmark detection unit 123...Landmark information storage unit 124…Landmark position correction unit 125…Self-map generation unit 126…Self-position estimation unit 127...Sensor cooperative error estimation unit 130…RAM 131…Landmark accumulation information 132…Landmark correction information 133…Self-generated map 140...Storage section 141…Self-generated map 150…Interface 160…ROM 170...Vehicle control device 171…Steering gear 172...Drive unit 173...braking device 201...sensor value acquisition unit 202...Odometry estimation unit 203…Landmark detection unit 204...Distance calculation unit 205...World coordinate calculation section 206…Landmark information storage unit 207…Landmark position correction unit 208…Self-map generation unit 209…Self-position estimation unit 220...Sensor cooperative error estimation unit 250…Self-map generation and self-location estimation unit

Claims

1. A self-position estimation device that generates a point cloud map from point cloud data of external environment information obtained by an on-board sensor mounted on a vehicle, and estimates a self-position, which is a position of the vehicle on the point cloud map, a landmark detection unit that detects landmarks from the point cloud map; a landmark information storage unit that stores observation position information of the detected landmarks; a landmark position correction unit that corrects the positions of the landmarks based on point cloud coordinates extracted from an error distribution detected when the observed position information of the landmarks is frequency-distributed, and calculates the importance of the landmarks based on the magnitude of the spread of the error distribution; an own map generating unit that stores a corrected map obtained by correcting the positions of the landmarks obtained by the landmark position correcting unit as a self-generated map; a self-position estimation unit that estimates the self-position of the vehicle currently traveling by weighting the stored self-generated map and the corrected map newly obtained from the landmark position correction unit while the vehicle is currently traveling by the importance of the weighted comparison.

2. The self-location estimation device according to claim 1 , The landmark position correction unit extracts, as point cloud coordinates, peak positions from an error distribution detected when the observed position information of the landmark is frequency-distributed, corrects the position of the landmark based on the peak positions, and calculates the importance of the landmark based on the magnitude of the spread of the error distribution for each peak position.

3. The self-location estimation device according to claim 1 , A self-position estimation device characterized in that the vehicle-mounted sensor is a distance sensor composed of multiple cameras, and the error distribution occurs in the line of sight direction due to parallax error of the distance sensor.

4. The self-location estimation device according to claim 3, The landmark position correction unit scans in a line of sight direction of the distance sensor and extracts a peak position of the error distribution as point cloud coordinates.

5. The self-location estimation device according to claim 1 , A self-location estimation device further comprising a sensor cooperative error estimation unit that estimates an amount of detection error of the landmark detected by a sensor other than the on-board sensor using the corrected map created based on the on-board sensor.

6. The self-location estimation device according to claim 1 , The self-location estimation device is characterized in that, in order to enable other vehicles to utilize the self-generated map, the self-generated map is stored in a server via a communication device, or the self-generated map generated by another vehicle is obtained via a communication device and can be utilized as the self-generated map of the vehicle itself.

7. The self-location estimation device according to claim 1 , The self-location estimation device is characterized in that the self-generated map is provided with absolute coordinates estimated from a GPS and odometry.

8. The self-location estimation device according to claim 1 , A self-location estimation device comprising a vehicle control device that automatically drives the vehicle in accordance with the self-generated map generated by the self-map generation unit and the self-location estimated by the self-location estimation unit provided in the self-location estimation device based on the self-generated map.

Images