Vehicle positioning device and vehicle positioning method

Abstract

To provide a vehicle location positioning device capable of increasing the accuracy of measuring the position of own vehicle based on sensor information from an on-vehicle sensor even in a situation where the position of objects on the ground around the vehicle cannot be detected.SOLUTION: A vehicle location positioning device 3 includes: a vehicle location positioning unit 12 that measures the position of own vehicle using satellite signals emitted from a satellite positioning system; and a position correct unit 16 that is configured to estimate a measurement error corresponding to the position of own vehicle measured by the vehicle location positioning unit 12 using an error estimation model for estimating the measurement error of the position in the vehicle location positioning unit 12 and correct the position of own vehicle measured by the vehicle location positioning unit 12 using the measurement error.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present disclosure relates to a vehicle positioning device and a vehicle positioning method. [Background technology]

[0002] There is a vehicle positioning device that measures the position of a vehicle (see, for example, Patent Document 1). The vehicle positioning device includes a coordinate acquisition unit, a recognition unit, and a control unit. The coordinate acquisition unit determines the position of the vehicle based on radio waves received from a satellite and acquires the position coordinates of the vehicle. The recognition unit acquires lane information indicating the lanes of the road on which the vehicle is traveling and recognizes the distance between the lane and the vehicle based on the lane information. The control unit corrects the position coordinates of the vehicle acquired by the coordinate acquisition unit based on the distance recognized by the recognition unit. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Laid-Open No. 2017-211193 Summary of the Invention [Problem to be solved by the invention]

[0004] In the vehicle positioning device disclosed in Patent Document 1, for example, when lanes are not marked on the road or when the lanes marked on the road are faded, the recognition unit may not be able to acquire lane information. The control unit has a problem in that, unless the recognition unit acquires lane information, it cannot improve the positioning accuracy by correcting the vehicle's position coordinates acquired by the coordinate acquisition unit.

[0005] The present disclosure has been made to solve the above-mentioned problems, and aims to provide a vehicle positioning device that can improve the positioning accuracy of the vehicle's position even in situations where it is not possible to detect the positions of features around the vehicle based on sensor information from on-board sensors. [Means for solving the problem]

[0006] The vehicle positioning device according to the present disclosure includes a vehicle positioning unit that uses satellite signals emitted from a satellite positioning system to determine the position of the vehicle, and a position correction unit that uses an error estimation model to estimate a positioning error of the vehicle position determined by the vehicle positioning unit, and corrects the vehicle position determined by the vehicle positioning unit using the positioning error. The vehicle navigation system includes a first feature position detection unit that acquires sensor information from an on-board sensor that observes the surroundings of the vehicle and detects the relative position of features existing around the vehicle relative to the vehicle based on the sensor information, a second feature position detection unit that detects the relative position of the features relative to the vehicle based on the position of the vehicle measured by the vehicle position measurement unit and map information, and an error calculation unit that calculates the error between the relative position detected by the first feature position detection unit and the relative position detected by the second feature position detection unit, wherein the error estimation model is a learned model that has learned the position measurement error of the position in the vehicle position measurement unit, and the position correction unit calculates the error between the signal of the position detection by the first feature position detection unit and the signal of the position detection by the vehicle position measurement unit. If the reliability is equal to or higher than the allowable reliability, learning data including position data indicating the position of the vehicle measured by the vehicle position measurement unit and error data indicating the error calculated by the error calculation unit is provided to a learning model, and the error indicated by the error data is used as a positioning error and the learning model learns the positioning error; if the calculated reliability is smaller than the allowable reliability, position data indicating the position of the vehicle measured by the vehicle position measurement unit is provided to the learning model, and positioning error data indicating the positioning error corresponding to the position indicated by the position data is obtained from the learning model, and the positioning error is used to correct the position of the vehicle measured by the vehicle position measurement unit. [Effects of the Invention]

[0007] According to the present disclosure, the accuracy of determining the position of the vehicle can be improved even in a situation where the positions of features existing around the vehicle cannot be detected based on sensor information from an on-board sensor. [Brief explanation of the drawings]

[0008] [Figure 1] 1 is a configuration diagram showing a vehicle position measuring device 3 according to a first embodiment. [Figure 2] 1 is a hardware configuration diagram showing hardware of a vehicle position measuring device 3 according to the first embodiment. [Figure 3] 1 is a hardware configuration diagram of a computer when the vehicle position measuring device 3 is realized by software, firmware, or the like. [Figure 4] 4 is a flowchart showing a vehicle positioning method, which is a processing procedure of the vehicle positioning device 3. [Figure 5] 10 is an explanatory diagram showing an example of detection of a white line by the first feature position detection unit 11. FIG. [Figure 6] 10 is an explanatory diagram showing an example of detection of a white line by the second feature position detection unit 13. FIG. [Figure 7] FIG. 7A is an explanatory diagram showing the relative positions detected by the first feature position detection unit 11, FIG. 7B is an explanatory diagram showing the relative positions detected by the second feature position detection unit 13, and FIG. 7C is an explanatory diagram showing the errors calculated by the error calculation unit 14. [Figure 8] FIG. 1 is an explanatory diagram showing a recurrent neural network that realizes a learning model 15a. [Figure 9] 1 is an explanatory diagram showing a travel section in which the first feature position detection unit 11 can detect the relative position of a feature, and a travel section in which the first feature position detection unit 11 cannot detect the relative position of a feature. FIG. [Figure 10] FIG. 10 is a configuration diagram showing a vehicle position measuring device 3 according to a second embodiment. [Figure 11] FIG. 10 is a hardware configuration diagram showing hardware of a vehicle position measuring device 3 according to a second embodiment. [Figure 12] FIG. 2 is an explanatory diagram showing an example of a state space model 17a. [Figure 13] FIG. 10 is a configuration diagram showing a vehicle position measuring device 3 according to a third embodiment. [Figure 14] FIG. 10 is a hardware configuration diagram showing hardware of a vehicle position measuring device 3 according to a third embodiment. [Figure 15] FIG. 2 is an explanatory diagram showing an example of an approximate function representing a positioning error Y corresponding to time t. [Figure 16] FIG. 10 is a configuration diagram showing a vehicle position measuring device 3 according to a fourth embodiment. [Figure 17] FIG. 10 is a hardware configuration diagram showing hardware of a vehicle position measuring device 3 according to a fourth embodiment. [Figure 18] 10 is an explanatory diagram showing an example of a positioning error calculated by a position correction unit 19. FIG. [Figure 19] FIG. 10 is a configuration diagram showing a vehicle position measuring device 3 according to a fifth embodiment. [Figure 20] FIG. 10 is a hardware configuration diagram showing hardware of a vehicle position measuring device 3 according to a fifth embodiment. [Figure 21] 3 is an explanatory diagram showing changes in the position of the vehicle measured by the vehicle position measuring unit 12. FIG. [Figure 22] FIG. 13 is a configuration diagram showing a vehicle position measuring device 3 according to a sixth embodiment. [Figure 23] FIG. 13 is a hardware configuration diagram showing hardware of a vehicle position measuring device 3 according to a sixth embodiment. [Figure 24] Fig. 24A is an explanatory diagram showing a driving environment condition where there are few obstacles, and Fig. 24B is an explanatory diagram showing a driving environment condition where there are many obstacles. DETAILED DESCRIPTION OF THE INVENTION

[0009] In order to explain the present disclosure in more detail, embodiments of the present disclosure will be described below with reference to the accompanying drawings.

[0010] Embodiment 1 FIG. 1 is a configuration diagram showing a vehicle position measuring device 3 according to the first embodiment. FIG. 2 is a hardware configuration diagram showing the hardware of the vehicle positioning device 3 according to the first embodiment. In FIG. 1, an on-vehicle sensor 1 includes, for example, a camera, a LiDAR (Light Detection and Ranging), a millimeter wave sensor, or a sonar. The on-board sensor 1 observes the surroundings of the vehicle and outputs sensor information indicating the observation results of the surroundings to the vehicle positioning device 3. The map information storage unit 2 is realized by, for example, a hard disk or a RAM (Random Access Memory). The map information storage unit 2 stores map information.

[0011] The vehicle position measuring device 3 includes a first feature position detecting unit 11, a vehicle position measuring unit 12, a second feature position detecting unit 13, an error calculating unit 14, a learning model storing unit 15, and a position correcting unit 16. The vehicle position measuring device 3 measures the position of the vehicle using sensor information from the on-board sensor 1 and satellite signals from the satellite positioning system.

[0012] The first feature position detection unit 11 is realized by, for example, a first feature position detection circuit 31 shown in FIG. The first feature position detection unit 11 acquires sensor information from the vehicle-mounted sensor 1. The first feature position detection unit 11 detects the relative positions of features around the vehicle relative to the vehicle (hereinafter referred to as "relative positions of features") based on sensor information. Examples of features include white lines on the road on which the vehicle is traveling, side walls of the road on which the vehicle is traveling, or signs and billboards installed on the road on which the vehicle is traveling. The first feature position detection unit 11 outputs the first feature position data to each of the error calculation unit 14 and the position correction unit 16 as position data indicating the relative position of the feature.

[0013] The vehicle position measuring unit 12 is realized by, for example, a vehicle position measuring circuit 32 shown in FIG. The vehicle position measuring unit 12 includes a receiver that receives satellite signals emitted from a satellite positioning system. The vehicle position measuring unit 12 measures the position of the vehicle using satellite signals. The vehicle position measuring unit 12 outputs the vehicle position data to the second feature position detecting unit 13 and the position correcting unit 16 as position data indicating the position of the vehicle.

[0014] The second feature position detection unit 13 is realized by, for example, a second feature position detection circuit 33 shown in FIG. The second feature position detection unit 13 acquires the vehicle position data from the vehicle position measurement unit 12 and acquires the map information from the map information storage unit 2 . The second feature position detection unit 13 detects the relative position of the feature based on the position of the vehicle indicated by the vehicle position data and the map information. The second feature position detection unit 13 outputs the second feature position data to the error calculation unit 14 as position data indicating the relative position of the feature.

[0015] The error calculation unit 14 is realized by, for example, the error calculation circuit 34 shown in FIG. The error calculation unit 14 calculates the error between the relative position of the feature detected by the first feature position detection unit 11 and the relative position of the feature detected by the second feature position detection unit 13 . The error calculation unit 14 outputs error data indicating the calculated error to the position correction unit 16.

[0016] The learning model storage unit 15 is realized by a learning model storage circuit 35 shown in FIG. The learning model storage unit 15 stores a learning model 15a as an error estimation model for estimating a position measurement error of the vehicle position measurement unit 12. The learning model 15a is realized by, for example, a recurrent neural network that can input time-series data. Examples of the recurrent neural network include a recurrent neural network (RNN) and a long short-term memory (LSTM). During learning of the learning model 15a, learning data is provided to the input layer of the recurrent neural network. The learning data includes the vehicle position data output from the vehicle position measurement unit 12 and the error data output from the error calculation unit 14. The error data is used as training data, and the learning model 15a learns the positioning error by taking the error indicated by the error data as the positioning error. When inferring the positioning error, the input layer of the recurrent neural network is given the vehicle position data output from the vehicle position measurement unit 12. As a result, the output layer of the recurrent neural network outputs positioning error data indicating the positioning error corresponding to the vehicle position data. Here, an example is shown in which learning model 15a is realized by a recurrent neural network, but this is merely an example, and learning model 15a may also be realized by a general neural network.

[0017] The position correction unit 16 is realized by, for example, a position correction circuit 36 ​​shown in FIG. The position correction unit 16 estimates a positioning error corresponding to the position of the vehicle measured by the vehicle position measurement unit 12, using an error estimation model for estimating the position measurement error of the position measured by the vehicle position measurement unit 12. Specifically, the position correction unit 16 provides the learning model 15a with vehicle position data indicating the position of the vehicle measured by the vehicle position measurement unit 12, and obtains from the learning model 15a positioning error data indicating the positioning error corresponding to the position indicated by the vehicle position data. The position correction unit 16 corrects the position of the vehicle measured by the vehicle position measurement unit 12 using the positioning error indicated by the positioning error data.

[0018] The position correction unit 16 may constantly correct the position of the vehicle using the positioning error, but as will be described below, the position correction unit 16 may correct the position of the vehicle using the positioning error only when the reliability R of the position detection by the first feature position detection unit 11 is smaller than the allowable reliability TR. In this case, the position correction unit 16 causes the learning model 15a to learn the positioning error if the reliability R is equal to or greater than the allowable reliability TR. Specifically, the position correction unit 16 calculates the reliability R of the position detection by the first feature position detection unit 11. If the reliability R of the position detection by the first feature position detection unit 11 is equal to or greater than the allowable reliability TR, the position correction unit 16 provides the learning data to the learning model 15a. The learning data includes vehicle position data indicating the position of the vehicle measured by the vehicle position measurement unit 12 and error data indicating the error calculated by the error calculation unit 14. The position corrector 16 regards the error indicated by the error data included in the learning data as a positioning error and causes the learning model 15a to learn the positioning error. If the reliability R of the position detection by the first feature position detection unit 11 is smaller than the allowable reliability TR, the position correction unit 16 provides the learning model 15a with vehicle position data indicating the position of the vehicle measured by the vehicle position measurement unit 12, and obtains from the learning model 15a positioning error data indicating the positioning error corresponding to the position indicated by the vehicle position data. The position correction unit 16 corrects the position of the vehicle measured by the vehicle position measurement unit 12 using the positioning error indicated by the positioning error data.

[0019] If the reliability R of the position detection by the first feature position detection unit 11 is equal to or greater than the allowable reliability TR, the position correction unit 16 outputs vehicle position data indicating the position of the vehicle measured by the vehicle position measurement unit 12 as the position measurement result, for example, to an autonomous driving system. If the reliability R of the position detection by the first feature position detection unit 11 is smaller than the allowable reliability TR, the position correction unit 16 outputs corrected position data indicating the corrected position as the positioning result to, for example, an automatic driving system. The automatic driving system controls the vehicle based on the position of the vehicle measured by the vehicle position measuring device 3 so that the vehicle travels within the road lane.

[0020] 1, it is assumed that each of the components of the vehicle positioning device 3, namely, the first feature position detection unit 11, the vehicle positioning unit 12, the second feature position detection unit 13, the error calculation unit 14, the learning model storage unit 15, and the position correction unit 16, is realized by dedicated hardware as shown in Fig. 2. That is, it is assumed that the vehicle positioning device 3 is realized by the first feature position detection circuit 31, the vehicle positioning circuit 32, the second feature position detection circuit 33, the error calculation circuit 34, the learning model storage circuit 35, and the position correction circuit 36. Here, the learning model memory circuit 35 may be, for example, a non-volatile or volatile semiconductor memory such as RAM, ROM (Read Only Memory), flash memory, EPROM (Erasable Programmable Read Only Memory), EEPROM (Electrically Erasable Programmable Read Only Memory), a magnetic disk, a flexible disk, an optical disk, a compact disk, a mini disk, or a DVD (Digital Versatile Disc). Furthermore, each of the first feature position detection circuit 31, the vehicle position measurement circuit 32, the second feature position detection circuit 33, the error calculation circuit 34 and the position correction circuit 36 ​​corresponds to, for example, a single circuit, a composite circuit, a programmed processor, a parallel programmed processor, an ASIC (Application Specific Integrated Circuit), an FPGA (Field-Programmable Gate Array), or a combination of these.

[0021] The components of the vehicle positioning device 3 are not limited to those realized by dedicated hardware, and the vehicle positioning device 3 may be realized by software, firmware, or a combination of software and firmware. Software or firmware is stored as a program in the memory of a computer. A computer refers to hardware that executes the program, such as a CPU (Central Processing Unit), central processing unit, processing unit, arithmetic unit, microprocessor, microcomputer, processor, or DSP (Digital Signal Processor).

[0022] FIG. 3 is a hardware configuration diagram of a computer in the case where the vehicle position measuring device 3 is realized by software, firmware, or the like. When the vehicle position measuring device 3 is realized by software, firmware, or the like, the learning model storage unit 15 is configured on the memory 51 of the computer. Programs for causing the computer to execute the respective processing procedures of the first feature position detecting unit 11, the vehicle position measuring unit 12, the second feature position detecting unit 13, the error calculating unit 14, and the position correcting unit 16 are stored in the memory 51. Then, the processor 52 of the computer executes the programs stored in the memory 51.

[0023] 2 shows an example in which each of the components of the vehicle positioning device 3 is realized by dedicated hardware, while Fig. 3 shows an example in which the vehicle positioning device 3 is realized by software, firmware, etc. However, this is merely an example, and some of the components in the vehicle positioning device 3 may be realized by dedicated hardware, and the remaining components may be realized by software, firmware, etc.

[0024] Next, the operation of the vehicle position measuring device 3 shown in FIG. 1 will be described. The on-board sensor 1 observes the surroundings of the vehicle and outputs sensor information indicating the observation results of the surroundings to the vehicle positioning device 3.

[0025] FIG. 4 is a flowchart showing a vehicle positioning method, which is a processing procedure of the vehicle positioning device 3. The first feature position detection unit 11 of the vehicle position measurement device 3 acquires sensor information from the vehicle-mounted sensor 1. The first feature position detection unit 11 detects the relative positions of features existing around the vehicle based on the sensor information (step ST1 in FIG. 4). When the first feature position detection unit 11 detects the relative position of a white line on a road relative to the vehicle as the relative position of a feature, the first feature position detection unit 11 detects the relative position of the white line, for example, by extracting the white line portion from an image captured by a camera included in the on-board sensor 1. Alternatively, the first feature position detection unit 11 detects the relative position of the white line by extracting the white line portion from reflection intensity information of a measurement point cloud of a LiDAR included in the on-board sensor 1. When the first feature position detection unit 11 detects the relative position of a side wall of a road relative to the vehicle as the relative position of a feature, for example, it detects the relative position of the side wall from a reflected wave from the side wall received by a millimeter wave sensor included in the on-board sensor 1. When the first feature position detection unit 11 detects the relative position of a signboard relative to the vehicle as the relative position of a feature, it detects the relative position of the signboard by, for example, performing image processing to identify the signboard that appears in the image captured by the camera included in the on-board sensor 1. The first feature position detection unit 11 outputs first feature position data indicating the relative position of the feature to each of the error calculation unit 14 and the position correction unit 16.

[0026] FIG. 5 is an explanatory diagram showing an example of detection of a white line by the first feature position detection unit 11. As shown in FIG. First, the first feature position detection unit 11 extracts the white line portion from the image captured by the camera. Next, the first feature position detection unit 11 performs coordinate conversion on the image captured by the camera into a bird's-eye view from an upper viewpoint. At this time, the first feature position detection unit 11 also performs coordinate conversion on the extracted white line portion, thereby projecting the white line portion onto a two-dimensional plane coordinate system. The two-dimensional plane coordinate system is a coordinate system that has a vehicle traveling direction and a vehicle lateral direction. The vehicle lateral direction is the vehicle width direction that is perpendicular to the vehicle traveling direction. White lines on roads are generally represented by dashed lines, so the first feature position detection unit 11 calculates the curved shape of a single white line by connecting multiple dashed white line segments. Next, the first feature position detection unit 11 identifies the position of the curved white line in the lateral direction of the host vehicle. The first feature position detection unit 11 calculates the distance from the host vehicle's position to the identified position of the white line. By calculating the curved shape of a single white line, the first feature position detection unit 11 can calculate the distance from the host vehicle's position to the position of the curve even if there is no white line in the lateral direction of the host vehicle.

[0027] The vehicle position measuring unit 12 receives satellite signals emitted from a satellite positioning system and measures the position of the vehicle using the satellite signals (step ST2 in FIG. 4). The vehicle position measuring unit 12 outputs vehicle position data indicating the position of the vehicle to the second feature position detecting unit 13 and the position correcting unit 16, respectively. The satellite positioning system is not limited to a system that serves satellite signals emitted from GPS satellites, but may also be a system that serves satellite signals emitted from quasi-zenith satellites, for example. 1, the vehicle position measuring unit 12 measures the position of the vehicle using satellite signals. However, this is merely an example, and the vehicle position measuring unit 12 may also obtain positioning data from electronic reference points installed by the Geospatial Information Authority of Japan and correct the position of the vehicle using the positioning data. The vehicle position measuring unit 12 may also measure the position of the vehicle using satellite signals emitted from a satellite positioning system and sensor signals from an inertial navigation sensor. Examples of inertial navigation sensors include an acceleration sensor or an angular velocity sensor. The vehicle position measuring unit 12 can calculate the speed and distance of the vehicle from an integrated value of the sensor signal from the acceleration sensor. The vehicle position measuring unit 12 can detect the traveling direction of the vehicle from the sensor signal from the angular velocity sensor. The vehicle position measuring unit 12 can determine the position of the vehicle by calculating the travel distance from an arbitrary point using the distance and traveling direction of the vehicle. The vehicle position measuring unit 12 may also calculate the travel distance of the vehicle by acquiring information indicating the vehicle speed or vehicle steering angle from a vehicle control unit (not shown).

[0028] The second feature position detection unit 13 acquires the vehicle position data from the vehicle position measurement unit 12 and acquires the map information from the map information storage unit 2 . The second feature position detection unit 13 detects the relative position of the feature based on the position of the vehicle indicated by the vehicle position data and the map information (step ST3 in FIG. 4). The second feature position detection unit 13 outputs second feature position data indicating the relative position of the feature to the error calculation unit 14. The feature detection process performed by the second feature position detection unit 13 will now be described in detail.

[0029] As shown in FIG. 6, the map information storage unit 2 stores three-dimensional shape data of high-precision maps as map information. FIG. 6 is an explanatory diagram showing an example of detection of a white line by the second feature position detection unit 13. In FIG. The second feature position detection unit 13 extracts the white line portion from the map information. The second feature position detection unit 13 performs coordinate conversion on the three-dimensional shape indicated by the map information into a bird's-eye view map from an upper viewpoint. At this time, the second feature position detection unit 13 also performs coordinate conversion on the extracted white line portion, thereby projecting the white line portion onto a two-dimensional plane coordinate system. The two-dimensional plane coordinate system is a coordinate system having a vehicle traveling direction and a vehicle lateral direction. The second feature position detection unit 13 calculates the curved shape of one white line by connecting a plurality of white line portions represented by dashed lines. Next, the second feature position detection unit 13 identifies the position of the curved white line in the lateral direction of the vehicle, and calculates the distance from the position of the vehicle to the identified position of the white line. 1, the second feature position detection unit 13 projects the white line portion onto a two-dimensional plane coordinate system. However, this is just one example, and the second feature position detection unit 13 may calculate the distance from the position of the vehicle to the position of the white line without converting the three-dimensional shape data indicating the white line portion into two-dimensional shape data.

[0030] The error calculation unit 14 acquires the first feature position data from the first feature position detection unit 11 and acquires the second feature position data from the second feature position detection unit 13 . The error calculation unit 14 calculates the error between the relative position of the feature indicated by the first feature position data and the relative position of the feature indicated by the second feature position data (step ST4 in FIG. 4). The error calculation unit 14 outputs error data indicating the calculated error to the position correction unit 16.

[0031] FIG. 7 is an explanatory diagram showing an example of calculation of the error by the error calculation unit 14. In FIG. 7A shows the relative positions detected by the first feature position detection unit 11, and FIG. 7B shows the relative positions detected by the second feature position detection unit 13. As shown in FIG. In FIG. 7A, the feature is a signboard, and V1 is a vector indicating the relative position of the signboard with respect to the vehicle. In FIG. 7B, the feature is a signboard, and V2 is a vector indicating the relative position of the signboard with respect to the vehicle. FIG. 7C shows the error calculated by the error calculation unit 14. V3 is a vector indicating the error between the relative position of the feature indicated by the first feature position data and the relative position of the feature indicated by the second feature position data. Vector V3 can be calculated, for example, by subtracting vector V1 from vector V2.

[0032] Here, the error calculation unit 14 calculates the error assuming that the feature is a sign. However, this is merely an example, and the error calculation unit 14 may calculate the error assuming that the feature is, for example, a white line or a side wall. When the error calculation unit 14 calculates the error assuming that the feature is, for example, a white line or a side wall, it may be difficult for the first feature position detection unit 11 to detect the relative position with respect to the vehicle traveling direction with high accuracy. For example, in a situation where there is no change in the shape of the road on which the host vehicle is traveling, it is difficult to detect the relative position with respect to the vehicle traveling direction with high accuracy. For this reason, the first feature position detection unit 11 may detect only the relative position in the lateral direction of the vehicle, and the error calculation unit 14 may calculate only the error in the lateral direction of the vehicle.

[0033] The position correction unit 16 calculates the reliability R of the position detection by the first feature position detection unit 11. An example of how the position corrector 16 calculates the reliability R will be described below. (1) If the first feature position detection unit 11 is unable to detect the relative position of the feature, the position correction unit 16 determines the reliability R of the position detection to be 0. If the reliability R is 0, the reliability R is a value smaller than the allowable reliability TR. (2) If the change in the relative position detected by the first feature position detection unit 11 within a certain period of time in the past is greater than a threshold, the position correction unit 16 determines the reliability R of the position detection to be a value smaller than the allowable reliability TR. If the change within the certain period of time in the past is equal to or smaller than the threshold, the position correction unit 16 determines the reliability R of the position detection to be a value equal to or larger than the allowable reliability TR. The threshold may be stored in an internal memory of the position correction unit 16 or may be provided from outside the vehicle positioning device 3. (3) When reliability information is included in the sensor information output from the vehicle-mounted sensor 1, the position correction unit 16 uses the reliability indicated by the reliability information as the reliability R of the position detection. (4) When the vehicle-mounted sensor 1 includes multiple types of sensor devices, and the sensor information indicated by the multiple sensor devices differs significantly, the position correction unit 16 determines the reliability R of the position detection to be 0.

[0034] The position correction unit 16 compares the reliability R of the position detection by the first feature position detection unit 11 with the allowable reliability TR. If the reliability R of the position detection is equal to or higher than the allowable reliability TR (step ST5 in FIG. 4: NO), the position correction unit 16 provides the learning data to the learning model 15a. The learning data includes vehicle position data indicating the position of the vehicle measured by the vehicle position measurement unit 12 and error data indicating the error calculated by the error calculation unit 14. The position corrector 16 regards the error indicated by the error data included in the learning data as a positioning error and causes the learning model 15a to learn the positioning error (step ST7 in FIG. 4). 1, the position correction unit 16 causes the learning model 15a to learn the positioning error. However, this is just one example, and the learning processing function of the position correction unit 16 may be implemented in an external server of the vehicle positioning device 3, and the external server may cause the learning model 15a to learn the positioning error. 1 includes a learning model storage unit 15. However, this is merely an example, and the learning model storage unit 15 may be provided outside the vehicle positioning device 3 or may be implemented in an external server.

[0035] FIG. 8 is an explanatory diagram showing a recurrent neural network that realizes the learning model 15a. A recurrent neural network recursively calculates the output result of the neural network by using, for example, the output result of the intermediate layer of the neural network calculated at time t as the input value of the intermediate layer of the neural network at time t + 1. In this way, the output result of the intermediate layer calculated at time t is passed on to the neural network at time t + 1, making it possible to learn using time-series data. Here, the learning data includes vehicle position data indicating the position of the vehicle measured by the vehicle position measurement unit 12 and error data indicating the error calculated by the error calculation unit 14. However, this is merely an example, and the position correction unit 16 may acquire vehicle speed information indicating the speed of the vehicle or steering angle information indicating the steering angle of the vehicle from a vehicle control unit (not shown), and include the vehicle speed information or steering angle information in the learning data. Furthermore, the position correction unit 16 may include relative position information indicating the relative position between the position of the vehicle measured by the vehicle position measurement unit 12 and the position of a satellite in the satellite positioning system in the learning data. In these cases, the learning model 15a learns the positioning error corresponding to the vehicle position data, the vehicle speed information, the steering angle information, or the relative position information.

[0036] FIG. 9 is an explanatory diagram showing a travel section in which the first feature position detection unit 11 can detect the relative position of a feature, and a travel section in which the first feature position detection unit 11 cannot detect the relative position of a feature. A travel section in which the relative position of a feature can be detected is a travel section in which the reliability R of position detection is equal to or greater than the allowable reliability TR. The driving section in which it is impossible to detect the relative position of a feature is not limited to a driving section in which the first feature position detection unit 11 is completely unable to detect the relative position of the feature, but also includes a driving section in which the reliability R of position detection is smaller than the allowable reliability TR.

[0037] If the reliability R of the position detection is smaller than the allowable reliability TR (step ST5: YES in Figure 4), the position correction unit 16 provides the learning model 15a with vehicle position data indicating the position of the vehicle measured by the vehicle position measurement unit 12, and obtains from the learning model 15a positioning error data indicating the positioning error corresponding to the position indicated by the vehicle position data. If the learning data includes vehicle speed information, steering angle information, or relative position information, the position correction unit 16 acquires the vehicle speed information, steering angle information, or relative position information and provides the vehicle speed information, steering angle information, or relative position information together with the host vehicle position data to the learning model 15a. Then, the position correction unit 16 acquires positioning error data from the learning model 15a, which indicates a positioning error corresponding to the host vehicle position data and the vehicle speed information, steering angle information, or relative position information. A section where the reliability R of position detection is smaller than the allowable reliability TR is a travel section where the relative position of a feature can be detected (see FIG. 9). The position correction unit 16 corrects the position of the vehicle measured by the vehicle position measurement unit 12 using the positioning error indicated by the positioning error data (step ST6 in FIG. 4).

[0038] The process of correcting the vehicle position by the position corrector 16 will now be described in detail. Based on the vehicle position data output from the vehicle position measurement unit 12, the position correction unit 16 sets the three-dimensional position of the vehicle as (x1, y1, z1). Furthermore, the position correction unit 16 calculates the three-dimensional position of the positioning error based on the positioning error data output from the learning model 15a as (x e ,y e ,z e ) as follows. The position correction unit 16 calculates the three-dimensional position (x1, y1, z1) of the vehicle from the three-dimensional position (x e ,y e ,z e ) is subtracted to obtain the corrected 3D position of the vehicle (x1-x e ,y1-y e ,z1-z e ) is calculated.

[0039] Here, the position correction unit 16 expresses the position of the host vehicle as a three-dimensional position (x1, y1, z1). However, this is only an example, and the position correction unit 16 converts the three-dimensional position (x1, y1, z1) into a two-dimensional position (x1, y1) in a two-dimensional coordinate system, and expresses the corrected two-dimensional position (x1-x e ,y1-y e The two-dimensional coordinate system is a coordinate system that has a vehicle traveling direction and a vehicle lateral direction. Furthermore, when the feature is a white line or a side wall, only the relative position in the lateral direction of the vehicle can be detected with high accuracy. Therefore, the position correction unit 16 converts the three-dimensional position (x1, y1, z1) into a one-dimensional position (x1) in the lateral direction of the vehicle, and calculates the corrected one-dimensional position (x1-x e ) may be calculated.

[0040] If the reliability R of the position detection by the first feature position detection unit 11 is equal to or greater than the allowable reliability TR (step ST5 in Figure 4: NO), the position correction unit 16 outputs vehicle position data indicating the position of the vehicle measured by the vehicle position measurement unit 12 as the position measurement result, for example, to an autonomous driving system (step ST8 in Figure 4). If the reliability R of the position detection by the first feature position detection unit 11 is smaller than the allowable reliability TR (step ST5 in Figure 4: YES), the position correction unit 16 outputs corrected position data indicating the corrected position as the positioning result to, for example, an automatic driving system (step ST8 in Figure 4). The automatic driving system controls the host vehicle based on the position of the host vehicle measured by the host vehicle position measuring device 3, for example, so that the host vehicle travels within the road lane.

[0041] In the above-described first embodiment, the vehicle positioning device 3 is configured to include a vehicle positioning unit 12 that uses satellite signals emitted from a satellite positioning system to position the position of the vehicle, and a position correction unit 16 that uses an error estimation model for estimating a positioning error in the vehicle positioning unit 12 to estimate a positioning error corresponding to the position of the vehicle positioned by the vehicle positioning unit 12, and corrects the position of the vehicle positioned by the vehicle positioning unit 12 using the positioning error. Therefore, the vehicle positioning device 3 can improve the positioning accuracy of the vehicle position based on sensor information from the on-board sensor 1 even in a situation where the positions of features existing around the vehicle cannot be detected.

[0042] In addition, in embodiment 1, the vehicle position measuring device 3 is configured to include a first feature position detection unit 11 that acquires sensor information from an on-board sensor 1 that observes the surroundings of the vehicle and detects the relative positions of features present around the vehicle with respect to the vehicle based on the sensor information, a second feature position detection unit 13 that detects the relative positions of features with respect to the vehicle based on the position of the vehicle measured by the vehicle position measuring unit 12 and map information, and an error calculation unit 14 that calculates the error between the relative position detected by the first feature position detection unit 11 and the relative position detected by the second feature position detection unit 13. Furthermore, in the vehicle position measuring device 3, the position correction unit 16 calculates the reliability of the position detection by the first feature position detection unit 11, and if the reliability is equal to or higher than the allowable reliability, provides the learning model 15a with position data indicating the position of the vehicle measured by the vehicle position measuring unit 12 and error data indicating the error calculated by the error calculation unit 14 as learning data, and causes the learning model 15a to learn the positioning error using the error indicated by the error data as a positioning error. Furthermore, if the calculated reliability is smaller than the allowable reliability, the position correction unit 16 provides the learning model 15a with position data indicating the position of the vehicle measured by the vehicle position measuring unit 12, obtains from the learning model 15a positioning error data indicating the positioning error corresponding to the position indicated by the position data, and corrects the position of the vehicle measured by the vehicle position measuring unit using the positioning error. Therefore, if the reliability of the position detection by the first feature position detection unit 11 is lower than the allowable reliability, the vehicle position measurement device 3 can improve the position measurement accuracy of the vehicle's position, while if the reliability of the position detection by the first feature position detection unit 11 is higher than the allowable reliability, the position measurement error can be learned by the learning model 15a, thereby improving the learning accuracy of the position measurement error.

[0043] Embodiment 2 In the second embodiment, a vehicle positioning device 3 including a position correction unit 18 that estimates a position measurement error in the vehicle position measurement unit 12 using a state space model as an error estimation model will be described.

[0044] Fig. 10 is a configuration diagram showing a vehicle position measuring device 3 according to embodiment 2. In Fig. 10, the same reference numerals as in Fig. 1 indicate the same or corresponding parts, and therefore description thereof will be omitted. Fig. 11 is a hardware configuration diagram showing the hardware of the vehicle position measuring device 3 according to embodiment 2. In Fig. 11, the same reference numerals as in Fig. 2 indicate the same or corresponding parts, and therefore a description thereof will be omitted. The vehicle position measuring device 3 shown in Figure 10 includes a first feature position detection unit 11, a vehicle position measuring unit 12, a second feature position detection unit 13, an error calculation unit 14, a state space model storage unit 17, and a position correction unit 18.

[0045] The state space model storage unit 17 is realized by a state space model storage circuit 37 shown in FIG. The state space model storage unit 17 stores a state space model 17a as an error estimation model. The state space model 17a is realized by, for example, a Kalman filter, a particle filter, or an α-β filter. The state space model 17 a represents a state equation of the position measurement error of the vehicle position measurement unit 12 .

[0046] The position correction unit 18 is realized by, for example, a position correction circuit 38 shown in FIG. The position corrector 18 calculates a positioning error by substituting the position of the host vehicle measured by the host vehicle position measuring unit 12 into the state equation indicated by the state space model 17a. The position correction unit 18 corrects the position of the vehicle measured by the vehicle position measurement unit 12 using the positioning error. The position correction unit 18 outputs corrected position data indicating the corrected position as a positioning result to, for example, an automatic driving system.

[0047] In the host vehicle position measuring device 3 shown in FIG. 10 , the position corrector 18 calculates the positioning error by substituting the position of the host vehicle measured by the host vehicle position measuring unit 12 into the state equation. However, this is merely an example, and the position corrector 18 may calculate the positioning error by substituting the relative position detected by the first feature position detecting unit 11 into the state equation. Furthermore, the position corrector 18 may calculate the positioning error by substituting the position of the host vehicle measured by the host vehicle position measuring unit 12 and the relative position detected by the first feature position detecting unit 11 into the state equation. Furthermore, the position corrector 18 may calculate the positioning error by substituting the error calculated by the error calculating unit 14 into the state equation.

[0048] 10, it is assumed that each of the components of the vehicle position measuring device 3, namely, the first feature position detecting unit 11, the vehicle position measuring unit 12, the second feature position detecting unit 13, the error calculating unit 14, the state space model storing unit 17, and the position correcting unit 18, is realized by dedicated hardware as shown in Fig. 11. In other words, it is assumed that the vehicle position measuring device 3 is realized by the first feature position detecting circuit 31, the vehicle position measuring circuit 32, the second feature position detecting circuit 33, the error calculating circuit 34, the state space model storing circuit 37, and the position correcting circuit 38. Here, the state space model storage circuit 37 corresponds to, for example, a nonvolatile or volatile semiconductor memory such as RAM, ROM, flash memory, EPROM, or EEPROM, a magnetic disk, a flexible disk, an optical disk, a compact disk, a minidisk, or a DVD. Furthermore, each of the first feature position detection circuit 31, the vehicle position measurement circuit 32, the second feature position detection circuit 33, the error calculation circuit 34 and the position correction circuit 38 corresponds to, for example, a single circuit, a composite circuit, a programmed processor, a parallel programmed processor, an ASIC, an FPGA, or a combination of these.

[0049] The components of the vehicle positioning device 3 are not limited to those realized by dedicated hardware, and the vehicle positioning device 3 may be realized by software, firmware, or a combination of software and firmware. When the vehicle position measuring device 3 is realized by software, firmware, or the like, the state space model storage unit 17 is configured on a memory 51 shown in Fig. 3. Programs for causing a computer to execute the respective processing procedures of the first feature position detecting unit 11, the vehicle position measuring unit 12, the second feature position detecting unit 13, the error calculating unit 14, and the position correcting unit 18 are stored in the memory 51. Then, a processor 52 shown in Fig. 3 executes the programs stored in the memory 51.

[0050] 11 shows an example in which each of the components of the vehicle positioning device 3 is realized by dedicated hardware, while Fig. 3 shows an example in which the vehicle positioning device 3 is realized by software, firmware, etc. However, this is merely an example, and some of the components in the vehicle positioning device 3 may be realized by dedicated hardware, and the remaining components may be realized by software, firmware, etc.

[0051] Next, the operation of the vehicle positioning device 3 shown in Fig. 10 will be described. Since the vehicle positioning device 3 is the same as that shown in Fig. 1 except for the state space model storage unit 17 and the position correction unit 18, the operation of the position correction unit 18 will mainly be described here. FIG. 12 is an explanatory diagram showing an example of the state space model 17a. In FIG. 12, the state value represents the state of the positioning error X. For example, X k-1 is the time t k-1 indicates the positioning error of the position of X k is the time t k This shows the positioning error of the position. The observed value Y represents the position of the vehicle measured by the vehicle position measuring unit 12. Y k-1 is the time T k-1The time t measured by the vehicle position measuring unit 12 at k-1 indicates the position of the vehicle, and Y k is the time t k-1 The position of the vehicle is shown.

[0052] If the observation noise is ON, the observation value Y is obtained by adding the observation noise ON to the positioning error X. Therefore, the observation equation is expressed as the following equation (1). X k =Y k -ON (1) Moreover, if it is assumed that the positioning error X varies at a constant speed, the state equation representing the change over time of the positioning error X is expressed as the following equation (2). X k =X k-1 +(dX k-1 / dt)·(t k -t k-1 )+PN (2) In equation (2), PN is the predicted noise. (dX k-1 / dt) is the positioning error X k-1 is the time derivative of

[0053] When calculating a state value at an arbitrary time from the observation equation and the state equation, if the state space model 17a is realized using a Kalman filter, the positioning error X k is expressed as the following equation (3). X k =X k-1 +G k ·(Y k -(X k-1 +dX k-1 / dt)·(t k -t k-1 )) (3) In equation (3), G k is the time t k is the Kalman gain at

[0054] The position correction unit 18 calculates the position of the vehicle measured by the vehicle position measurement unit 12 at time t k Vehicle position Y kBy substituting into equation (3), time t k Positioning error X k Calculate. The position correction unit 18 calculates the positioning error X k is used to correct the position of the vehicle measured by the vehicle position measurement unit 12. The position correction process is similar to the position correction process by the position correction unit 16 shown in Fig. 1, and therefore a detailed description thereof will be omitted. The position correction unit 18 outputs corrected position data indicating the corrected position as a positioning result to, for example, an automatic driving system.

[0055] In the second embodiment described above, the error estimation model is a state space model that indicates a state equation of a positioning error of the position in the vehicle position measuring unit 12. The vehicle position measuring device 3 shown in FIG. 10 is configured so that the position correction unit calculates a positioning error by substituting the position of the vehicle measured by the vehicle position measuring unit 12 into the state equation and corrects the position of the vehicle measured by the vehicle position measuring unit 12 using the positioning error. Therefore, similar to the vehicle position measuring device 3 shown in FIG. 1, the vehicle position measuring device 3 shown in FIG. 10 can improve the positioning accuracy of the vehicle position based on sensor information from the on-board sensor 1 even in a situation where the positions of features existing around the vehicle cannot be detected based on the sensor information from the on-board sensor 1. Furthermore, the vehicle position measuring device 3 shown in FIG. 10 can calculate a positioning error without learning a large amount of learning data in advance, as with the learning model 15a shown in FIG. 1.

[0056] Embodiment 3 In the third embodiment, a vehicle positioning device 3 will be described that includes a position correction unit 19 that calculates a positioning error by substituting the time at which the error is calculated by the error calculation unit 14 into an approximation function.

[0057] Fig. 13 is a configuration diagram showing a vehicle position measuring device 3 according to embodiment 3. In Fig. 13, the same reference numerals as in Fig. 1 indicate the same or corresponding parts, and therefore description thereof will be omitted. Fig. 14 is a hardware configuration diagram showing the hardware of the vehicle position measuring device 3 according to embodiment 3. In Fig. 14, the same reference numerals as in Fig. 2 indicate the same or corresponding parts, and therefore description thereof will be omitted. The vehicle position measuring device 3 shown in Figure 13 includes a first feature position detection unit 11, a vehicle position measuring unit 12, a second feature position detection unit 13, an error calculation unit 14, a learning model storage unit 15, and a position correction unit 19.

[0058] The position correction unit 19 is realized by, for example, a position correction circuit 39 shown in FIG. The position correction process by the position correction unit 19 is similar to the position correction process by the position correction unit 16 shown in FIG. The position correction unit 19 calculates the reliability R of the position detection by the first feature position detection unit 11, similar to the position correction unit 16 shown in FIG. If the reliability R of the position detection by the first feature position detection unit 11 is equal to or greater than the allowable reliability TR, the position correction unit 19 calculates the positioning error of the position in the vehicle position measurement unit 12 by substituting the time at which the error was calculated by the error calculation unit 14 into an approximation function that represents the positioning error corresponding to the time. The position correction unit 19 provides the learning model 15a with learning data including the subject vehicle position data indicating the position of the subject vehicle measured by the subject vehicle position measurement unit 12 and error data indicating the calculated positioning error. The position correction unit 19 causes the learning model 15a to learn the positioning error indicated by the error data included in the learning data.

[0059] 13, it is assumed that each of the components of the vehicle positioning device 3, namely, the first feature position detection unit 11, the vehicle positioning unit 12, the second feature position detection unit 13, the error calculation unit 14, the learning model storage unit 15, and the position correction unit 19, is realized by dedicated hardware as shown in Fig. 14. In other words, it is assumed that the vehicle positioning device 3 is realized by the first feature position detection circuit 31, the vehicle positioning circuit 32, the second feature position detection circuit 33, the error calculation circuit 34, the learning model storage circuit 35, and the position correction circuit 39. Each of the first feature position detection circuit 31, the vehicle position measurement circuit 32, the second feature position detection circuit 33, the error calculation circuit 34 and the position correction circuit 39 corresponds to, for example, a single circuit, a composite circuit, a programmed processor, a parallel programmed processor, an ASIC, an FPGA, or a combination thereof.

[0060] The components of the vehicle positioning device 3 are not limited to those realized by dedicated hardware, and the vehicle positioning device 3 may be realized by software, firmware, or a combination of software and firmware. When the vehicle position measuring device 3 is realized by software, firmware, or the like, the learning model storage unit 15 is configured on a memory 51 shown in Fig. 3. Programs for causing a computer to execute the respective processing procedures of the first feature position detecting unit 11, the vehicle position measuring unit 12, the second feature position detecting unit 13, the error calculating unit 14, and the position correcting unit 19 are stored in the memory 51. Then, a processor 52 shown in Fig. 3 executes the programs stored in the memory 51.

[0061] 14 shows an example in which each of the components of the vehicle positioning device 3 is realized by dedicated hardware, while Fig. 3 shows an example in which the vehicle positioning device 3 is realized by software, firmware, etc. However, this is merely an example, and some of the components in the vehicle positioning device 3 may be realized by dedicated hardware, and the remaining components may be realized by software, firmware, etc.

[0062] Next, the operation of the vehicle positioning device 3 shown in Fig. 13 will be described. Since the vehicle positioning device 3 is the same as that shown in Fig. 1 except for the position corrector 19, only the operation of the position corrector 19 will be described here. The internal memory of the position correction unit 19 stores an approximate expression of an approximate function such as the following equation (4). Y=αt 3 +βt 2 +γt+δ (4) In equation (4), Y is the positioning error corresponding to time t, and α, β, γ, and δ are coefficients of a polynomial that represents an approximate function. FIG. 15 is an explanatory diagram showing an example of an approximate function representing the positioning error Y corresponding to the time t. In FIG. 15, the horizontal axis represents time t, and the vertical axis represents positioning error Y. In FIG. 15, the following two approximate expressions are shown as examples of approximate expressions of the approximate functions. Y=α1t 3 +β1t 2 +γ1t+δ1(4') Y=α2t 3 +β2t 2 +γ2t+δ2(4”)

[0063] The position correction unit 19 calculates the coefficients α, β, γ, and δ of the polynomial in advance. Specifically, the position correction unit 19 substitutes each of the errors Y at the multiple times t calculated by the error calculation unit 14 into the polynomial shown in equation (4) to create multiple equations with different values ​​of the error Y. Then, the position correction unit 19 calculates the coefficients α, β, γ, and δ of the polynomial by solving the multiple equations as simultaneous equations.

[0064] When the first feature position detection unit 11 is able to detect the relative position of a feature in a travel section, the position correction unit 19 calculates the positioning error Y by substituting the time t at which the error was calculated by the error calculation unit 14 into equation (4). As described above, a travel section in which the relative position of a feature can be detected is a travel section in which the reliability R of position detection by the first feature position detection unit 11 is equal to or greater than the allowable reliability TR. The position correction unit 19 provides the learning model 15a with learning data including vehicle position data indicating the position of the vehicle measured by the vehicle position measurement unit 12 and error data indicating the positioning error Y, thereby causing the learning model 15a to learn the positioning error indicated by the error data included in the learning data.

[0065] In the third embodiment described above, the vehicle positioning device 3 shown in FIG. 13 is configured such that the position correction unit 19 calculates the positioning error of the vehicle positioning unit 12 by substituting the time at which the error is calculated by the error calculation unit 14 into an approximation function representing the positioning error corresponding to time, and provides learning data including position data indicating the vehicle's position measured by the vehicle positioning unit 12 and error data indicating the calculated positioning error to the learning model 15a, thereby causing the learning model 15a to learn the positioning error. Therefore, similar to the vehicle positioning device 3 shown in FIG. 1, the vehicle positioning device 3 shown in FIG. 13 can improve the positioning accuracy of the vehicle's position based on sensor information from the on-board sensor 1 even in a situation where the positions of features around the vehicle cannot be detected based on the sensor information from the on-board sensor 1. Furthermore, similar to the vehicle positioning device 3 shown in FIG. 1, the vehicle positioning device 3 shown in FIG. 13 can cause the learning model 15a to learn the positioning error.

[0066] Embodiment 4 In the fourth embodiment, a vehicle positioning device 3 is described that includes a position correction unit 20 that calculates the average value of errors calculated at multiple times by an error calculation unit 14 as the positioning error of the vehicle position measurement unit 12.

[0067] Fig. 16 is a configuration diagram showing a vehicle position measuring device 3 according to embodiment 4. In Fig. 16, the same reference numerals as in Fig. 1 indicate the same or corresponding parts, and therefore description thereof will be omitted. Fig. 17 is a hardware configuration diagram showing the hardware of the vehicle position measuring device 3 according to embodiment 4. In Fig. 17, the same reference numerals as in Fig. 2 indicate the same or corresponding parts, and therefore description thereof will be omitted. The vehicle position measuring device 3 shown in Figure 16 includes a first feature position detection unit 11, a vehicle position measuring unit 12, a second feature position detection unit 13, an error calculation unit 14, a learning model storage unit 15, and a position correction unit 20.

[0068] The position correction unit 20 is realized by, for example, a position correction circuit 40 shown in FIG. The position correction process by the position correction unit 20 is similar to the position correction process by the position correction unit 16 shown in FIG. The position correction unit 20 calculates the reliability R of the position detection by the first feature position detection unit 11, similar to the position correction unit 16 shown in FIG. If the reliability R of the position detection by the first feature position detection unit 11 is equal to or greater than the allowable reliability TR, the position correction unit 20 calculates the average value of the errors at multiple times calculated by the error calculation unit 14 as the positioning error of the position in the vehicle position positioning unit 12. The position correction unit 20 provides the learning model 15a with learning data including the subject vehicle position data indicating the subject vehicle position measured by the subject vehicle position measurement unit 12 and the calculated average value. The position corrector 19 causes the learning model 15a to learn the average value contained in the learning data.

[0069] 16, it is assumed that each of the components of the vehicle positioning device 3, namely, the first feature position detection unit 11, the vehicle positioning unit 12, the second feature position detection unit 13, the error calculation unit 14, the learning model storage unit 15, and the position correction unit 20, is realized by dedicated hardware as shown in Fig. 17. In other words, it is assumed that the vehicle positioning device 3 is realized by the first feature position detection circuit 31, the vehicle positioning circuit 32, the second feature position detection circuit 33, the error calculation circuit 34, the learning model storage circuit 35, and the position correction circuit 40. Each of the first feature position detection circuit 31, the vehicle position measurement circuit 32, the second feature position detection circuit 33, the error calculation circuit 34 and the position correction circuit 40 corresponds to, for example, a single circuit, a composite circuit, a programmed processor, a parallel programmed processor, an ASIC, an FPGA, or a combination thereof.

[0070] The components of the vehicle positioning device 3 are not limited to those realized by dedicated hardware, and the vehicle positioning device 3 may be realized by software, firmware, or a combination of software and firmware. When the vehicle position measuring device 3 is realized by software, firmware, or the like, the learning model storage unit 15 is configured on a memory 51 shown in Fig. 3. Programs for causing a computer to execute the respective processing procedures of the first feature position detecting unit 11, the vehicle position measuring unit 12, the second feature position detecting unit 13, the error calculating unit 14, and the position correcting unit 40 are stored in the memory 51. Then, a processor 52 shown in Fig. 3 executes the programs stored in the memory 51.

[0071] 17 shows an example in which each of the components of the vehicle positioning device 3 is realized by dedicated hardware, while Fig. 3 shows an example in which the vehicle positioning device 3 is realized by software, firmware, etc. However, this is merely an example, and some of the components in the vehicle positioning device 3 may be realized by dedicated hardware, and the remaining components may be realized by software, firmware, etc.

[0072] Next, the operation of the vehicle positioning device 3 shown in Fig. 16 will be described. Since the vehicle positioning device 3 is the same as that shown in Fig. 1 except for the position correcting unit 20, only the operation of the position correcting unit 20 will be described here. FIG. 18 is an explanatory diagram showing an example of the positioning error calculated by the position corrector 19. As shown in FIG. In FIG. 18, the horizontal axis represents time, and the vertical axis represents the error calculated by the error calculation unit 14.

[0073] The position correction unit 20 calculates the reliability R of the position detection by the first feature position detection unit 11, similar to the position correction unit 16 shown in FIG. If the reliability R of the position detection by the first feature position detection unit 11 is equal to or greater than the allowable reliability TR, the position correction unit 20 calculates the average value of the errors at multiple times calculated by the error calculation unit 14 as the positioning error of the position in the vehicle position positioning unit 12. For example, the position correction unit 20 k When calculating the positioning error of time t k The time span W up to a certain time before kThe average value of the multiple positioning errors included in The position correction unit 20 provides the learning model 15a with learning data including the subject vehicle position data indicating the subject vehicle position measured by the subject vehicle position measurement unit 12 and the calculated average value. The position corrector 19 causes the learning model 15a to learn the average value contained in the learning data.

[0074] In the fourth embodiment described above, the vehicle positioning device 3 shown in FIG. 16 is configured such that the position correction unit 20 calculates an average value of errors calculated by the error calculation unit 14 at multiple times as the positioning error of the position in the vehicle positioning unit 12, and provides learning data including position data indicating the position of the vehicle measured by the vehicle positioning unit 12 and the average value to the learning model 15a, causing the learning model 15a to learn the average value. Therefore, similar to the vehicle positioning device 3 shown in FIG. 1, the vehicle positioning device 3 shown in FIG. 16 can improve the positioning accuracy of the vehicle position even in a situation where the positions of features existing around the vehicle cannot be detected based on sensor information from the on-board sensor 1. Furthermore, the vehicle positioning device 3 shown in FIG. 16 can suppress short-term fluctuations in the positioning error included in the learning data, thereby enabling more stable learning of the positioning error than the vehicle positioning device 3 shown in FIG. 1.

[0075] Embodiment 5 In embodiment 5, a vehicle positioning device 3 is described that is equipped with a position correction unit 21 that, when the change in the position of the vehicle measured by the vehicle position measuring unit 12 is greater than or equal to a threshold, reduces the learning weight for the position of the vehicle measured by the vehicle position measuring unit 12 before the change in position becomes greater than or equal to the threshold when having the learning model 15a learn the positioning error.

[0076] Fig. 19 is a configuration diagram showing a vehicle position measuring device 3 according to embodiment 5. In Fig. 19, the same reference numerals as in Fig. 1 indicate the same or corresponding parts, and therefore description thereof will be omitted. Fig. 20 is a hardware configuration diagram showing the hardware of the vehicle positioning device 3 according to embodiment 5. In Fig. 20, the same reference numerals as in Fig. 2 indicate the same or corresponding parts, and therefore description thereof will be omitted. The vehicle position measuring device 3 shown in Figure 19 includes a first feature position detection unit 11, a vehicle position measuring unit 12, a second feature position detection unit 13, an error calculation unit 14, a learning model storage unit 15, and a position correction unit 21.

[0077] The position correction unit 21 is realized by, for example, a position correction circuit 41 shown in FIG. The position correction process by the position correction unit 21 is similar to the position correction process by the position correction unit 16 shown in FIG. The position correction unit 21 calculates the reliability R of the position detection by the first feature position detection unit 11, similar to the position correction unit 16 shown in FIG. Similar to the position correction unit 16 shown in FIG. 1, the position correction unit 21 provides the learning model 15a with learning data including vehicle position data indicating the position of the vehicle measured by the vehicle position measurement unit 12 and error data indicating the error calculated by the error calculation unit 14. Similar to the position corrector 16 shown in FIG. 1, the position corrector 21 regards the error indicated by the error data included in the learning data as the positioning error and causes the learning model 15a to learn the positioning error. If the change in the position of the host vehicle measured by the host vehicle position measuring unit 12 is equal to or greater than a threshold, the position corrector 21 lowers the learning weight for the position of the host vehicle measured by the host vehicle position measuring unit 12 before the change in the position becomes equal to or greater than the threshold, compared to the learning weight for the position of the host vehicle measured by the host vehicle position measuring unit 12 after the change in the position becomes equal to or greater than the threshold, when making the learning model 15a learn the positioning error. The threshold may be stored in an internal memory of the position corrector 21, or may be provided from outside the host vehicle position measuring device 3.

[0078] 19, it is assumed that each of the components of the vehicle positioning device 3, namely, the first feature position detection unit 11, the vehicle positioning unit 12, the second feature position detection unit 13, the error calculation unit 14, the learning model storage unit 15, and the position correction unit 21, is realized by dedicated hardware as shown in Fig. 20. In other words, it is assumed that the vehicle positioning device 3 is realized by the first feature position detection circuit 31, the vehicle positioning circuit 32, the second feature position detection circuit 33, the error calculation circuit 34, the learning model storage circuit 35, and the position correction circuit 41. Each of the first feature position detection circuit 31, the vehicle position measurement circuit 32, the second feature position detection circuit 33, the error calculation circuit 34 and the position correction circuit 41 corresponds to, for example, a single circuit, a composite circuit, a programmed processor, a parallel programmed processor, an ASIC, an FPGA, or a combination thereof.

[0079] The components of the vehicle positioning device 3 are not limited to those realized by dedicated hardware, and the vehicle positioning device 3 may be realized by software, firmware, or a combination of software and firmware. When the vehicle position measuring device 3 is realized by software, firmware, or the like, the learning model storage unit 15 is configured on a memory 51 shown in Fig. 3. Programs for causing a computer to execute the respective processing procedures of the first feature position detecting unit 11, the vehicle position measuring unit 12, the second feature position detecting unit 13, the error calculating unit 14, and the position correcting unit 21 are stored in the memory 51. Then, a processor 52 shown in Fig. 3 executes the programs stored in the memory 51.

[0080] 20 shows an example in which each of the components of the vehicle positioning device 3 is realized by dedicated hardware, while Fig. 3 shows an example in which the vehicle positioning device 3 is realized by software, firmware, etc. However, this is merely an example, and some of the components in the vehicle positioning device 3 may be realized by dedicated hardware, and the remaining components may be realized by software, firmware, etc.

[0081] Next, the operation of the vehicle positioning device 3 shown in Fig. 19 will be described. Since the vehicle positioning device 3 is the same as that shown in Fig. 1 except for the position corrector 21, only the operation of the position corrector 21 will be described here. FIG. 21 is an explanatory diagram showing changes in the position of the vehicle measured by the vehicle position measuring unit 12. As shown in FIG. 21, the horizontal axis represents time, and the vertical axis represents the amount of change in the position of the vehicle measured by the vehicle position measuring unit 12.

[0082] The position correction unit 21 monitors changes in the position of the host vehicle measured by the host vehicle position measurement unit 12. If the position of the host vehicle changes suddenly, the amount of change in position may exceed a threshold, as shown in FIG. When the position of the host vehicle changes suddenly, the behavior of the positioning error before the sudden change in the position of the host vehicle often differs from the behavior of the positioning error after the sudden change in the position of the host vehicle. Therefore, when the learning model 15a learns the positioning error, it is desirable to reduce the influence of the position of the host vehicle measured by the host vehicle position measuring unit 12 before the position of the host vehicle changes suddenly.

[0083] The position correction unit 21 determines the time when the amount of change in position becomes equal to or greater than the threshold value as t k If so, when the learning model 15a learns the positioning error, k-1 At time t k+1 In this case, learning is performed so that the weight is lower than the weight of learning for the position of the vehicle measured by the vehicle position measuring unit 12. As a result, in the learning of the learning model 15a, at time t k-1 The influence of the position of the host vehicle on the time domain is reduced.

[0084] In the above-described fifth embodiment, the vehicle positioning device 3 shown in FIG. 19 is configured such that, when the change in the position of the host vehicle measured by the vehicle positioning unit 12 is equal to or greater than a threshold, the position correction unit 21 causes the learning model 15a to learn the positioning error by lowering the learning weight for the position of the host vehicle measured by the vehicle positioning unit 12 before the change in the position becomes equal to or greater than the threshold, compared to the learning weight for the position of the host vehicle measured by the vehicle positioning unit 12 after the change in the position becomes equal to or greater than the threshold. Therefore, similar to the vehicle positioning device 3 shown in FIG. 1, the vehicle positioning device 3 shown in FIG. 19 can improve the positioning accuracy of the host vehicle position even in a situation where the positions of features existing around the host vehicle cannot be detected based on sensor information from the on-board sensor 1. Furthermore, the vehicle positioning device 3 shown in FIG. 19 can reduce the impact on the learning of the learning model 15a even if the position of the host vehicle measured by the vehicle positioning unit 12 changes suddenly.

[0085] In the vehicle position measuring device 3 shown in FIG. 19, the position correcting unit 21 causes the learning model 15a to learn the positioning error at time t k-1 In this example, the learning weight for the position of the vehicle measured by the vehicle position measuring unit 12 is reduced. However, this is merely an example, and the position correction unit 21 may reduce the number of position data indicating the position of the vehicle measured by the vehicle position measuring unit 12 before the change in position becomes equal to or greater than a threshold when making the learning model 15a learn the positioning error. When the error estimation model is the state space model 17a as in the vehicle position measuring device 3 shown in FIG. 10, the time when the amount of change in position becomes equal to or greater than the threshold value is t k If so, the position correction unit 18 calculates the state value X k The Kalman gain G k By increasing the value of , the influence of the state equation is reduced, and the observed value Y k The weight of the value of may be decreased. In the vehicle position measuring device 3 shown in FIG. 13, the time when the amount of change in position becomes equal to or greater than the threshold value is t k If so, the position correction unit 19 k Alternatively, the positioning error at a time earlier than the predetermined time may not be used. In the vehicle position measuring device 3 shown in FIG. 16, the time when the amount of change in position becomes equal to or greater than the threshold value is t k If so, the position correction unit 20 k Alternatively, the average value may be calculated without using the positioning errors at times earlier than the current time.

[0086] Embodiment 6 In the sixth embodiment, a vehicle position measuring device 3 including a surrounding environment detection unit 22 that detects the surrounding environment of the vehicle will be described.

[0087] Fig. 22 is a configuration diagram showing a vehicle position measuring device 3 according to embodiment 6. In Fig. 22, the same reference numerals as in Fig. 1 indicate the same or corresponding parts, and therefore description thereof will be omitted. Fig. 23 is a hardware configuration diagram showing the hardware of the vehicle position measuring device 3 according to embodiment 6. In Fig. 23, the same reference numerals as in Fig. 2 indicate the same or corresponding parts, and therefore a description thereof will be omitted. The vehicle position measuring device 3 shown in Figure 22 includes a first feature position detection unit 11, a vehicle position measuring unit 12, a second feature position detection unit 13, an error calculation unit 14, a learning model memory unit 15, a surrounding environment detection unit 22, and a position correction unit 23.

[0088] The ambient environment detection unit 22 is realized by, for example, an ambient environment detection circuit 42 shown in FIG. The surrounding environment detection unit 22 acquires sensor information from the on-board sensor 1, and detects the surrounding environment of the vehicle based on the sensor information. The surrounding environment detection unit 22 outputs the detection result of the surrounding environment to the position correction unit 23.

[0089] The position correction unit 23 is realized by, for example, a position correction circuit 43 shown in FIG. The position correction process by the position correction unit 23 is similar to the position correction process by the position correction unit 16 shown in FIG. The position correction unit 23 calculates the reliability R of the position detection by the first feature position detection unit 11, similar to the position correction unit 16 shown in FIG. Similar to the position correction unit 16 shown in FIG. 1, the position correction unit 23 provides the learning model 15a with learning data including vehicle position data indicating the position of the vehicle measured by the vehicle position measurement unit 12 and error data indicating the error calculated by the error calculation unit 14. Similar to the position corrector 16 shown in FIG. 1, the position corrector 23 regards the error indicated by the error data included in the learning data as the positioning error and causes the learning model 15a to learn the positioning error. When the position correction unit 23 causes the learning model 15a to learn the positioning error, the position correction unit 23 changes the learning weight for the error calculated by the error calculation unit 14 based on the detection result of the surrounding environment by the surrounding environment detection unit 22.

[0090] 22, it is assumed that each of the components of the vehicle position measuring device 3, namely, the first feature position detecting unit 11, the vehicle position measuring unit 12, the second feature position detecting unit 13, the error calculating unit 14, the learning model storing unit 15, the surrounding environment detecting unit 22, and the position correcting unit 23, is realized by dedicated hardware as shown in Fig. 23. In other words, it is assumed that the vehicle position measuring device 3 is realized by the first feature position detecting circuit 31, the vehicle position measuring circuit 32, the second feature position detecting circuit 33, the error calculating circuit 34, the learning model storing circuit 35, the surrounding environment detecting circuit 42, and the position correcting circuit 43. Each of the first feature position detection circuit 31, the vehicle position measurement circuit 32, the second feature position detection circuit 33, the error calculation circuit 34, the surrounding environment detection circuit 42 and the position correction circuit 43 may be, for example, a single circuit, a composite circuit, a programmed processor, a parallel programmed processor, an ASIC, an FPGA, or a combination thereof.

[0091] The components of the vehicle positioning device 3 are not limited to those realized by dedicated hardware, and the vehicle positioning device 3 may be realized by software, firmware, or a combination of software and firmware. When the vehicle position measuring device 3 is realized by software, firmware, or the like, the learning model storage unit 15 is configured on a memory 51 shown in Fig. 3. Programs for causing a computer to execute the respective processing procedures of the first feature position detecting unit 11, the vehicle position measuring unit 12, the second feature position detecting unit 13, the error calculating unit 14, the surrounding environment detecting unit 22, and the position correcting unit 23 are stored in the memory 51. Then, a processor 52 shown in Fig. 3 executes the programs stored in the memory 51.

[0092] 23 shows an example in which each of the components of the vehicle positioning device 3 is realized by dedicated hardware, while Fig. 3 shows an example in which the vehicle positioning device 3 is realized by software, firmware, etc. However, this is merely an example, and some of the components in the vehicle positioning device 3 may be realized by dedicated hardware, and the remaining components may be realized by software, firmware, etc.

[0093] Next, the operation of the vehicle positioning device 3 shown in Fig. 22 will be described. Except for the surrounding environment detection unit 22 and the position correction unit 23, the vehicle positioning device 3 is the same as that shown in Fig. 1. Therefore, here, the operation of the surrounding environment detection unit 22 and the position correction unit 23 will be described. Fig. 24A is an explanatory diagram showing a driving environment condition where there are few obstacles, and Fig. 24B is an explanatory diagram showing a driving environment condition where there are many obstacles. In the example of FIG. 24A, there are no obstacles, such as buildings, between the satellite and the vehicle that would block the satellite signal emitted from the satellite, so the driving environment conditions are such that there are few obstacles. In the example of FIG. 24B, there are obstacles such as buildings between the satellite and the vehicle that block the satellite signal emitted from the satellite, resulting in a driving environment condition with many obstacles.

[0094] The surrounding environment detection unit 22 acquires sensor information from the on-board sensor 1, and detects the surrounding environment of the vehicle based on the sensor information. The surrounding environment detection unit 22 outputs the detection result of the surrounding environment to the position correction unit 23. The detection result of the surrounding environment indicates whether or not there is an obstacle, such as a building, between the satellite and the vehicle that blocks the satellite signal emitted from the satellite. If the satellite signal emitted from the satellite is, for example, a GPS (Global Positioning System) signal, the surrounding environment detection unit 22 can confirm the three-dimensional position of the satellite by referring to the GPS signal. The surrounding environment detection unit 22 draws a line connecting the three-dimensional position of the satellite and the three-dimensional position of the vehicle indicated by the vehicle position data output from the vehicle position positioning unit 12, and if there is a building or the like blocking the line, it determines that there is an obstacle between the satellite and the vehicle. If there is no building or the like blocking the line segment, the surrounding environment detection unit 22 determines that there is no obstacle between the satellite and the vehicle. The surrounding environment detection unit 22 acquires attribute information of buildings and the like existing around the vehicle by referring to the map information stored in the map information storage unit 2. The attribute information is information indicating the size, shape, color, and the like of the buildings and the like. The surrounding environment detection unit 22 can confirm the three-dimensional position of the space occupied by the building or the like by referring to the attribute information of the building or the like.

[0095] When the position correction unit 23 causes the learning model 15a to learn the positioning error, the position correction unit 23 changes the learning weight for the error calculated by the error calculation unit 14 based on the detection result of the surrounding environment by the surrounding environment detection unit 22. Specifically, if there are obstacles around the vehicle, the positioning error of the vehicle position measuring unit 12 may be large, and as a result, the accuracy of the error calculation by the error calculation unit 14 may be degraded. Therefore, when the detection result of the surrounding environment by the surrounding environment detection unit 22 indicates that an obstacle is present, the position correction unit 23 changes the learning weight for the error calculated by the error calculation unit 14 so that it is smaller than the learning weight for the error when the detection result does not indicate that an obstacle is present. This reduces the influence of radio wave interference caused by obstacles around the vehicle during learning of the learning model 15a.

[0096] In the sixth embodiment described above, the vehicle positioning device 3 shown in FIG. 22 is configured to include a surrounding environment detection unit 22 that acquires sensor information from the on-board sensor 1 and detects the environment surrounding the vehicle based on the sensor information. Furthermore, in the vehicle positioning device 3 shown in FIG. 22, when the position correction unit 23 causes the learning model 15a to learn the positioning error, the position correction unit 23 changes the learning weight for the error calculated by the error calculation unit 14 based on the detection result of the surrounding environment by the surrounding environment detection unit 22. Therefore, similar to the vehicle positioning device 3 shown in FIG. 1, the vehicle positioning device 3 shown in FIG. 22 can improve the positioning accuracy of the vehicle position even in a situation where the positions of features present around the vehicle cannot be detected based on the sensor information from the on-board sensor 1. Furthermore, the vehicle positioning device 3 shown in FIG. 22 can reduce the influence of radio wave interference on the learning of the learning model 15a when an obstacle is present around the vehicle.

[0097] In the vehicle position measuring device 3 according to the first to sixth embodiments, the position corrector 16 and the like correct the position of the vehicle in the traveling direction and the position of the vehicle in the width direction as the position of the vehicle. However, this is merely an example, and the position correction unit 16 and the like may correct only the position of the host vehicle in the vehicle width direction. For example, in a situation where there is no change in the shape of the road on which the host vehicle is traveling, it may be difficult to detect the relative position with respect to the vehicle traveling direction with high accuracy. In such a case, the first feature position detection unit 11 detects only the relative position in the vehicle lateral direction, which is the vehicle width direction, and the error calculation unit 14 calculates only the error in the vehicle lateral direction. Then, the position correction unit 16 and the like correct only the position of the host vehicle in the vehicle width direction.

[0098] In addition, the present disclosure allows for free combination of the respective embodiments, modification of any of the components of the respective embodiments, or omission of any of the components of the respective embodiments. [Explanation of symbols]

[0099] REFERENCE SIGNS LIST 1 on-vehicle sensor, 2 map information storage unit, 3 vehicle positioning device, 11 first feature position detection unit, 12 vehicle position measurement unit, 13 second feature position detection unit, 14 error calculation unit, 15 learning model storage unit, 15a learning model, 16 position correction unit, 17 state space model storage unit, 17a state space model, 18 position correction unit, 19 position correction unit, 20 position correction unit, 21 position correction unit, 22 surrounding environment detection unit, 23 position correction unit, 31 first feature position detection circuit, 32 vehicle position measurement circuit, 33 second feature position detection circuit, 34 error calculation circuit, 35 learning model storage circuit, 36 position correction circuit, 37 state space model storage circuit, 38 position correction circuit, 39 position correction circuit, 40 position correction circuit, 41 position correction circuit, 42 surrounding environment detection circuit, 43 Position correction circuit, 51 memory, 52 processor.

Claims

1. a vehicle positioning unit that uses satellite signals emitted from a satellite positioning system to measure the position of the vehicle; a position correction unit that estimates a positioning error corresponding to the position of the host vehicle measured by the host vehicle position measurement unit using an error estimation model for estimating a position measurement error in the host vehicle position measurement unit, and corrects the position of the host vehicle measured by the host vehicle position measurement unit using the positioning error; a first feature position detection unit that acquires sensor information from an on-board sensor that observes the surroundings of the host vehicle, and detects relative positions of features present around the host vehicle with respect to the host vehicle based on the sensor information; a second feature position detection unit that detects a relative position of the feature with respect to the vehicle based on the position of the vehicle measured by the vehicle position measurement unit and map information; an error calculation unit that calculates an error between the relative position detected by the first feature position detection unit and the relative position detected by the second feature position detection unit, the error estimation model is a learning model in which the positioning error of the position measured by the vehicle position measurement unit has been learned, The position correction unit calculating a reliability of the position detection by the first feature position detection unit, and if the reliability is equal to or greater than an allowable reliability, providing learning data to the learning model, the learning data including position data indicating the position of the vehicle measured by the vehicle position measurement unit and error data indicating the error calculated by the error calculation unit, and causing the learning model to learn the positioning error, with the error indicated by the error data being considered as the positioning error; If the calculated reliability is smaller than the allowable reliability, the position data indicating the position of the host vehicle measured by the host vehicle position measurement unit is provided to the learning model, and positioning error data indicating the positioning error corresponding to the position indicated by the position data is obtained from the learning model, and the position of the host vehicle measured by the host vehicle position measurement unit is corrected using the positioning error. Vehicle positioning device.

2. The position correction unit 2. The vehicle positioning device according to claim 1, wherein, if the calculated reliability is equal to or greater than the allowable reliability, the position data indicating the position of the vehicle measured by the vehicle position measuring unit is output as the positioning result, and if the calculated reliability is smaller than the allowable reliability, the position data indicating the corrected position is output as the positioning result.

3. the error estimation model is a state space model that indicates a state equation of a position measurement error in the vehicle position measurement unit, The position correction unit 2. The vehicle positioning device according to claim 1, wherein the positioning error is calculated by substituting the position of the vehicle measured by the vehicle positioning unit into the state equation, and the positioning error is used to correct the position of the vehicle measured by the vehicle positioning unit.

4. The position correction unit The vehicle positioning device according to claim 1, characterized in that the positioning error of the position in the vehicle positioning unit is calculated by substituting the time at which the error is calculated by the error calculation unit into an approximation function representing the positioning error corresponding to time, and learning data including the position data indicating the position of the vehicle measured by the vehicle positioning unit and the error data indicating the calculated positioning error is provided to the learning model, thereby allowing the learning model to learn the positioning error.

5. The position correction unit The vehicle positioning device according to claim 1, characterized in that the positioning error of the position in the vehicle positioning unit is calculated as an average value of the errors at multiple times calculated by the error calculation unit, and learning data including the position data indicating the position of the vehicle measured by the vehicle positioning unit and the average value is provided to the learning model, and the average value is learned by the learning model.

6. The position correction unit The vehicle positioning device of claim 1, characterized in that when the change in the position of the vehicle measured by the vehicle positioning unit is greater than or equal to a threshold, when the positioning error is learned into the learning model, the learning weight for the position of the vehicle measured by the vehicle positioning unit before the change in position becomes greater than or equal to the threshold is lowered below the learning weight for the position of the vehicle measured by the vehicle positioning unit after the change in position becomes greater than or equal to the threshold.

7. a surrounding environment detection unit that acquires the sensor information from the on-board sensor and detects an environment around the host vehicle based on the sensor information; The position correction unit The vehicle positioning device according to claim 1, characterized in that when the positioning error is learned into the learning model, the learning weight for the error calculated by the error calculation unit is changed based on the detection results of the surrounding environment by the surrounding environment detection unit.

8. The position correction unit 2. The vehicle positioning device according to claim 1, wherein the error estimation model is used to estimate the positioning error in a vehicle width direction, which is a direction perpendicular to the traveling direction of the vehicle, as the positioning error corresponding to the position of the vehicle measured by the vehicle positioning unit, and the position of the vehicle in the vehicle width direction measured by the vehicle positioning unit is corrected using the positioning error in the vehicle width direction.

9. The position correction unit 2. The vehicle position measurement device according to claim 1, further comprising: acquiring vehicle speed information indicating a speed of the vehicle; steering angle information indicating a steering angle of the vehicle; or relative position information indicating a relative position between the position of the vehicle measured by the vehicle position measurement unit and a position of a satellite in the satellite positioning system; using the error estimation model, estimating the positioning error corresponding to the position of the vehicle measured by the vehicle position measurement unit and the speed indicated by the vehicle speed information, the steering angle indicated by the steering angle information, or the relative position indicated by the relative position information; and correcting the position of the vehicle measured by the vehicle position measurement unit using the positioning error.

10. a vehicle position measuring unit for measuring the position of the vehicle using a satellite signal emitted from a satellite positioning system; a position correction unit uses an error estimation model for estimating a position measurement error of the vehicle position measurement unit to estimate a position measurement error corresponding to the vehicle position measured by the vehicle position measurement unit, and corrects the vehicle position measured by the vehicle position measurement unit using the position measurement error; a first feature position detection unit that acquires sensor information from an on-board sensor that observes the surroundings of the host vehicle, and detects relative positions of features present around the host vehicle with respect to the host vehicle based on the sensor information; a second feature position detection unit that detects a relative position of the feature with respect to the vehicle based on the position of the vehicle measured by the vehicle position measurement unit and map information; an error calculation unit calculates an error between the relative position detected by the first feature position detection unit and the relative position detected by the second feature position detection unit; A vehicle positioning method, comprising: the error estimation model is a learning model in which the positioning error of the position measured by the vehicle position measurement unit has been learned, The position correction unit calculating a reliability of the position detection by the first feature position detection unit, and if the reliability is equal to or greater than an allowable reliability, providing learning data to the learning model, the learning data including position data indicating the position of the vehicle measured by the vehicle position measurement unit and error data indicating the error calculated by the error calculation unit, and causing the learning model to learn the positioning error, with the error indicated by the error data being considered as the positioning error; If the calculated reliability is smaller than the allowable reliability, the position data indicating the position of the host vehicle measured by the host vehicle position measurement unit is provided to the learning model, and positioning error data indicating the positioning error corresponding to the position indicated by the position data is obtained from the learning model, and the position of the host vehicle measured by the host vehicle position measurement unit is corrected using the positioning error. Vehicle positioning method.

Images