JP2019190994A - Method and device for estimating self-position - Google Patents

Abstract

To estimate the self-position of a vehicle precisely when there are errors of a traveling path or the position of a target.SOLUTION: The method for estimating a self-position according to the present invention is for estimating the self-position of an own vehicle using a traveling path model and a target model. The method includes: calculating an error of the travel path model and an error of the target model; and determining which of the result of calculation output from a travel path model and the result of calculation output from the target model to use for starting the estimation of the target position according to the size of the error of the travel path model and the error of the target model.SELECTED DRAWING: Figure 10

Description

  The present invention relates to a self-position estimation method and a self-position estimation apparatus that estimate a self-position of a host vehicle using a travel locus model and a target model.

  Conventionally, a method for correcting map data stored in a car navigation device or the like is known. The invention described in Patent Document 1 compares travel locus data using GPS with map data, and corrects a shift in map data with respect to the travel locus data. Thereby, the position of the target on the map is corrected, and the self position of the vehicle is calculated using the corrected position of the target.

JP 2007-310198 A

  Here, in the conventional apparatus mentioned above, map data is correct | amended with respect to travel locus data, and it is assumed that there is no error in travel locus data. However, there is a problem that the own position of the own vehicle cannot be accurately estimated because the travel locus data actually includes an error.

  Therefore, the present invention has been made in view of the above problems, and an object of the present invention is to provide a self-position estimation method and a self-position estimation apparatus that can accurately estimate the self-position of the host vehicle.

  A self-position estimation method and apparatus according to an aspect of the present invention calculate an error of a travel locus model and an error of a target model. Then, depending on the magnitude of the error in the travel trajectory model and the target model, either the calculation result output from the travel trajectory model or the calculation result output from the target model is used. Decide whether to start estimating target information.

  According to the present invention, it is possible to accurately estimate the own position of the host vehicle.

FIG. 1 is a block diagram showing the configuration of the self-position estimation apparatus according to the first embodiment of the present invention. FIG. 2 is a diagram for explaining a travel locus model used in the self-position estimation apparatus according to the first embodiment of the present invention. FIG. 3 is a diagram for explaining a target model used in the self-position estimation apparatus according to the first embodiment of the present invention. FIG. 4 is a flowchart showing a processing procedure of trajectory correction processing by the self-position estimation apparatus according to the first embodiment of the present invention. FIG. 5 is a diagram for explaining a trajectory correction method by the self-position estimation apparatus according to the first embodiment of the present invention. FIG. 6 is a diagram for explaining a trajectory correction method by the self-position estimation apparatus according to the first embodiment of the present invention. FIG. 7 is a flowchart showing a processing procedure of target position correction processing by the self-position estimation apparatus according to the first embodiment of the present invention. FIG. 8 is a diagram for explaining a method for correcting a target position by the self-position estimation apparatus according to the first embodiment of the present invention. FIG. 9 is a diagram for explaining a method for correcting a target position by the self-position estimation apparatus according to the first embodiment of the present invention. FIG. 10 is a flowchart showing a processing procedure of self-position estimation processing by the self-position estimation apparatus according to the first embodiment of the present invention. FIG. 11 is a diagram for explaining an error of a white line recorded in conventional map information. FIG. 12 is a diagram for explaining a model determination method used when starting self-position estimation by the self-position estimation apparatus according to the second embodiment of the present invention. FIG. 13 is a flowchart showing a processing procedure of self-position estimation processing by the self-position estimation apparatus according to the second embodiment of the present invention.

[First Embodiment]
Hereinafter, a first embodiment of the present invention will be described with reference to the drawings. In the description of the drawings, the same portions are denoted by the same reference numerals, and description thereof is omitted.

[Configuration of self-position estimation device]
With reference to FIG. 1, the structure of the self-position estimation apparatus which concerns on this embodiment is demonstrated. As shown in FIG. 1, the self-position estimation device 1 includes a camera 2, a laser range finder 3, a GPS receiver 4, a gyro sensor 5, a map database 6, a controller 10, and a communication device 7. The self-position estimation apparatus 1 is mounted on a vehicle, and the mounted vehicle is a vehicle that has an automatic driving function and performs driving support.

  The camera 2 includes an image sensor such as a charge-coupled device (CCD) or a complementary metal oxide semiconductor (CMOS). The camera 2 is mounted on the host vehicle and photographs the surroundings of the host vehicle. The camera 2 has an image processing function and detects a target or the like from the photographed image. A target is an object provided on a road or a sidewalk, such as a traffic light, a utility pole, a traffic sign, or the like. The target also includes a road structure. The road structure is a component of the road, and includes, for example, a road edge, a curb, a median strip, and the like in addition to a road dividing line and a stop line such as a white line and a yellow line. The camera 2 outputs the detected data to the controller 10. The target may be any object that is stationary, and may be, for example, a building or a characteristic part of a building.

  The laser range finder 3 is a sensor that detects a target around the host vehicle. Specifically, the laser range finder 3 scans the laser beam within a certain angle range, receives the reflected light at that time, and detects the time difference between the time of laser emission and the time of receiving the reflected light. The laser range finder 3 detects the relative distance and direction of the target with respect to the host vehicle, and outputs the detected data to the controller 10. The laser range finder 3 is installed around a bonnet, a bumper, a license plate, a headlight, a side mirror, and the like.

  The GPS receiver 4 detects the current location of the vehicle on the ground by receiving radio waves from an artificial satellite. The GPS receiver 4 outputs the detected data to the controller 10.

  The gyro sensor 5 detects the yaw rate (rotational angular velocity) around the vertical axis of the center of gravity of the host vehicle, and outputs the detected data to the controller 10.

  The self-position estimation device 1 is connected to a sensor group (not shown) mounted on the vehicle. For example, it is connected to an accelerator sensor, a steering sensor, a brake sensor, a vehicle speed sensor, an acceleration sensor, a wheel speed sensor, etc., and sensor values output from these sensor groups can be acquired.

  The map database 6 is a database stored in a car navigation device or the like, and stores various data necessary for route guidance such as target information and facility information. The map database 6 is a high-precision map, and stores target information including road structure information such as the number of road lanes and road boundaries. The map database 6 outputs map information to the controller 10 in response to a request from the controller 10. Information such as target data, travel locus data, trajectory data, and the like is also stored.

  The various data such as the target data are not necessarily limited to those acquired from the map database 6, and may be acquired by a sensor included in the own vehicle. Moreover, you may acquire using vehicle-to-vehicle communication and road-to-vehicle communication. For example, when various types of data such as target data are stored in the server 25, the controller 10 can acquire these data at any time by communication. Further, the controller 10 can periodically obtain the latest map information from the server 25 and update the held map information. In the present embodiment, the map database 6 includes a first map (a high-precision map, a map used for vehicle route guidance), and a second map may be included to supplement the first map. . This second map is created by the vehicle or the external server 25 when the first map alone is incomplete or when more detailed map information is required.

  The controller 10 is a circuit that processes data acquired from the camera 2, the laser range finder 3, the GPS receiver 4, the gyro sensor 5, and the map database 6, and is configured by, for example, an IC or an LSI. When the controller 10 grasps this functionally, the controller 10 classifies the travel locus model creation unit 11, the target model creation unit 12, the trajectory calculation unit 13, the target position calculation unit 14, and the trajectory correction unit 15. be able to. Further, the controller 10 can be classified into a target calculation unit 16, a target position correction unit 17, a model error determination unit 18, a matching unit 19, and a self-position estimation unit 20.

  The controller 10 includes a general-purpose electronic circuit including a microcomputer, a microprocessor, and a CPU, and peripheral devices such as a memory. And it operates as the self-position estimation apparatus 1 by executing a specific program. Each function of the controller 10 can be implemented by one or a plurality of processing circuits. The processing circuit includes a programmed processing device such as, for example, a processing device including an electrical circuit, and an application specific integrated circuit (ASIC) or conventional circuit arranged to perform the functions described in the embodiments. It also includes devices such as parts.

  The travel locus model creation unit 11 creates a travel locus model using the travel data of the host vehicle. When the travel trajectory data obtained from the odometry of the host vehicle is input, the travel trajectory model adjusts the travel trajectory data obtained from the odometry using a predetermined model and outputs the adjusted result as trajectory data of the host vehicle. To do. The odometry is the amount of movement per unit time of the vehicle, and can be calculated using various sensor values obtained from a sensor group mounted on the vehicle. In addition, as shown in FIG. 2, the travel locus model is composed of a plurality of points Xi (i = 1 to n) on the travel locus in a predetermined section (section surrounded by a one-dot chain line shown in FIG. 2). . Each point Xi is composed of coordinates (x, y) and a vehicle direction θ (traveling direction).

  The traveling locus model creation unit 11 acquires a large number of coordinates (x, y) and vehicle orientation θ from the traveling locus of the host vehicle, learns and models the traveling locus of each point Xi using the acquired traveling data, A travel locus model is created so that the travel locus can be reproduced. In addition, the travel locus model is generated by learning the travel locus of the host vehicle before executing the traveling support of the host vehicle. The predetermined section is a section divided by a predetermined traveling time or traveling scene. The travel time is, for example, 3 to 4 seconds. The traveling scene is, for example, a scene from the entrance to the exit of the intersection, a right turn or a left turn scene. The reason for dividing by the travel time or the travel scene in this way is that when the travel time is long or the travel scene is complicated, various data enters and the accuracy of the travel locus model may be lowered.

  The traveling locus model creation unit 11 can create a traveling locus model with high accuracy by creating a traveling locus model in a section divided by a predetermined traveling time or traveling scene. The travel locus model creation unit 11 does not necessarily need to use the travel data of the host vehicle, and may use travel data of other vehicles other than the host vehicle collected in the server 25 without using the travel data in advance. A created model may be used. In addition, the travel locus model creation unit 11 may create a model before executing the travel support of the host vehicle, and use travel data obtained in the process of executing the travel support of the host vehicle. A model may be created. Furthermore, the traveling locus model creation unit 11 may update the model in real time using traveling data obtained in the traveling process.

  Specifically, the travel locus model creation unit 11 obtains an average position for each point Xi in a predetermined section from the acquired data. Next, the travel locus model creation unit 11 calculates a variance-covariance matrix from the acquired travel data, and obtains a principal component vector. At this time, the traveling locus model creation unit 11 obtains the number that can sufficiently represent the traveling locus as the number of principal component vectors. Next, the traveling locus model creation unit 11 arranges the principal component vectors in a column and creates a matrix A. The matrix A is a trajectory control coefficient. The travel locus model creation unit 11 creates a travel locus model X by adding the average position of each point of the model and the matrix A and the control parameter q of the matrix A. The travel locus model X is expressed as follows using the equations (1) and (2).

  The traveling locus model creation unit 11 may store the created traveling locus model X in the server 25 or may be stored in the host vehicle. Note that only one traveling locus model X or a plurality of traveling locus models X may be generated. The predetermined section may be divided by a time such as 5 seconds or 10 seconds, or may be divided by an operation. As an example in the case of dividing by operation, a period from entering the intersection to exiting is defined as one section. The traveling locus model creation unit 11 may store the traveling locus model X in another vehicle using inter-vehicle communication.

  The target model creation unit 12 creates a target model using target data obtained by detecting targets around the host vehicle. When target data related to a target obtained by a sensor of the host vehicle is input, the target model adjusts the target data obtained by the sensor using a predetermined model, and the adjusted result is displayed around the subject vehicle. Output as target data. As shown in FIG. 3, the target model is composed of a plurality of points Li (i = 1 to n) on a white line in a predetermined section (a section surrounded by a one-dot chain line shown in FIG. 3). In FIG. 3, a white line is used as an example of the target. Each point Li is composed of coordinates (x, y) and a white line direction φ.

  The target model creation unit 12 acquires a large number of coordinates (x, y) and white line direction φ from the target data, learns and models the white line of each point Li using the acquired data, and can reproduce the white line. A target model is created so that In addition, the target model is generated by learning the target around the host vehicle before executing driving support of the host vehicle. Note that the point Li = (x, y) is the position of the i-th white line obtained by dividing the white line in a certain section at intervals of 10 cm, for example.

  Specifically, the target model creation unit 12 obtains an average position for each point Li in a predetermined section from the acquired data. Next, the target model creation unit 12 calculates a variance-covariance matrix from the acquired data to obtain a principal component vector. At this time, the target model creation unit 12 obtains the number that can sufficiently express the target as the number of principal component vectors. Next, the target model creating unit 12 creates a matrix B by arranging the principal component vectors in a column. The matrix B is a trajectory control coefficient. The target model creation unit 12 creates the target model L by adding the average position of each point of the model and the control parameter p of the matrix B and the matrix B. The target model L is expressed as follows using the equations (3) and (4), similarly to the traveling locus models of the equations (1) and (2).

  The target model creation unit 12 may store the created target model L in the server 25 or in the host vehicle. Only one target model L or a plurality of target models L may be generated. Further, the predetermined section may be divided by a time such as 5 seconds or 10 seconds. The target model creation unit 12 may store the target model L in another vehicle using inter-vehicle communication. In addition, the target model creation unit 12 may create a model before executing the driving support of the host vehicle, and uses target data obtained in the process of executing the driving support of the host vehicle. You may create a model. Furthermore, the target model creation unit 12 may update the model in real time using target data obtained in the course of traveling.

  Here, since the vehicle travels along the white line on the road, there is a correlation between the travel locus and the shape of the white line, which is an approximation. Therefore, there is also a correlation between the travel locus model and the target model, which is an approximation. Therefore, there is a correlation between the control coefficient A of the travel locus model and the control coefficient B of the target model, and there is also a correlation between the control parameter q of the travel locus model and the control parameter p of the target model.

  The trajectory calculation unit 13 calculates trajectory data using the travel locus model X stored in the server 25. The trajectory is a travel locus of the host vehicle calculated using the travel locus model X. The trajectory calculation unit 13 initializes the vehicle orientation θ and the control parameter q with predetermined values (for example, 0), and inputs them to the travel locus model X to calculate the trajectory. The control parameter q may be initialized with a value calculated backward from the odometry trajectory. The traveling locus of the host vehicle can also be obtained using odometry. Odometry is a method for determining the moving distance and direction of a moving body according to the rotation angle and rotation angular velocity of a tire. However, when odometry is used, there is a possibility that an error due to tire slip or the like cannot be avoided, and there is a possibility that the estimation accuracy of the travel locus is lowered. In the present embodiment, since a trajectory using the travel locus model X is used, it is possible to suppress a decrease in the estimation accuracy of the travel locus.

  The trajectory calculation unit 13 may calculate the trajectory before acquiring the target using odometry. That is, the trajectory data may be calculated based on the travel locus data obtained by odometry of the host vehicle. Further, the trajectory calculation unit 13 may acquire a target using a trajectory calculated from odometry, or may calculate a trajectory after acquiring a target using a travel locus model.

  The target position calculation unit 14 calculates the position of the target on the map using the trajectory calculated by the trajectory calculation unit 13. The target position calculation unit 14 acquires the distance from each point on the trajectory to the target, and acquires the direction θ of the host vehicle when the distance is acquired. Each point may be a plurality of positions, and it is only necessary to acquire the distance to the target and the direction of the vehicle a plurality of times even at the same position. The target position calculation unit 14 calculates the position of the target using the distance to the target and the vehicle orientation θ.

  The trajectory correction unit 15 obtains a difference between the position of the target calculated by the target position calculation unit 14 and the position of the target recorded in the map information, and estimates this difference as an error of the travel locus model. Then, the trajectory correction unit 15 corrects the trajectory so that the position of the target calculated by the target position calculation unit 14 matches the position of the target recorded in the map information. For example, the trajectory is corrected by using an optimization method such as the Gauss-Newton method and a method in which the vehicle orientation θ and the control parameter q are adjusted and gradually approached to the true value in the repeated calculation.

  The trajectory correction unit 15 gives the distance from the calculated target position to the target acquired at each point and the direction of the vehicle as constraint conditions, and corrects the trajectory using the travel locus model X. Thereby, even if there is an error in the position of each point where the distance to the target and the direction of the vehicle are acquired (for example, GPS error or odometry error), the trajectory correction unit 15 can accurately calculate the trajectory. it can. As a result, the self-position of the vehicle can be accurately calculated. Further, the trajectory correction unit 15 can calculate the trajectory more accurately by repeating this process.

  Further, the trajectory correction unit 15 obtains the variance of the target position calculated from each point on the trajectory, and corrects the trajectory by adjusting the vehicle orientation θ and the control parameter q when the variance is larger than the threshold. Also good. The variance is an index indicating the degree of dispersion and can be obtained using a general method. If the variance is larger than the threshold value, it means that there is a lot of data far from the average value, and the variance is large. On the other hand, when the variance is less than or equal to the threshold value, it means that there is a lot of data close to the average value and the variance is small. In other words, when the variance is equal to or less than the threshold value, it means that the position of the target can be estimated with high accuracy. On the other hand, if the variance is larger than the threshold value, it means that the position of the target has not been accurately estimated. The threshold value may be obtained through experiments or simulations, or may be set arbitrarily. Then, the trajectory correction unit 15 corrects the trajectory by repeatedly adjusting the vehicle orientation θ and the control parameter q until the variance becomes equal to or less than the threshold value.

  When the trajectory correction unit 15 determines that the calculated position of the target coincides with the position of the target recorded in the map information as a result of performing the trajectory correction described above, the trajectory correction ends. . When using an optimization method such as the Gauss-Newton method as a method for determining whether or not they match, if it is determined that the value has converged and approached the true value, it is recorded in the calculated target position and map information. It is determined that the position of the target has been matched. In the case of using variance, when the variance is equal to or smaller than the threshold value, it is determined that the calculated target position matches the target position recorded in the map information.

  The target calculation unit 16 calculates a target including a road structure using the target model L stored in the server 25. The target calculation unit 16 initializes the vehicle orientation θ and the control parameter p with predetermined values (for example, 0), and inputs them to the target model L to calculate the target. Further, the target calculation unit 16 may calculate target data around the host vehicle based on the target data obtained by the sensor of the host vehicle. Then, the target calculation unit 16 arranges the calculated target data on the map on the basis of the travel locus data obtained by the odometry of the host vehicle, and calculates the position of the target.

  The target position correction unit 17 obtains a difference between the position of the target calculated by the target calculation unit 16 and the position of the target recorded in the map information, and estimates this difference as an error of the target model. The target position correcting unit 17 corrects the target data so that the target position calculated by the target calculating unit 16 matches the target position recorded in the map information. For example, by using an optimization method such as the Gauss-Newton method, the target data is corrected using a method of gradually approaching the true value in the repeated calculation while adjusting the vehicle orientation θ and the control parameter p. To do. The target position correcting unit 17 gives the trajectory as a constraint condition and corrects the target data using the target model L.

  Further, the target position correcting unit 17 may obtain the variance of the target position, and correct the target position by adjusting the control parameter p when the variance is larger than the threshold value. The threshold value may be obtained through experiments or simulations, or may be set arbitrarily. Then, the target position correcting unit 17 corrects the position of the target by repeatedly adjusting the control parameter p until the variance becomes equal to or less than the threshold value. Further, the position of the target may be corrected by adjusting both the vehicle orientation θ and the control parameter p.

  When the target position correction unit 17 determines that the calculated target position matches the target position recorded in the map information as a result of correcting the target position described above, The correction of the target position is completed. When using an optimization method such as the Gauss-Newton method as a method for determining whether or not they match, if it is determined that the value has converged and approached the true value, it is recorded in the calculated target position and map information. It is determined that the position of the target has been matched. In the case of using variance, when the variance is equal to or smaller than the threshold value, it is determined that the calculated target position matches the target position recorded in the map information.

  The model error determination unit 18 uses either the calculation result output from the travel trajectory model or the calculation result output from the target model according to the magnitude of the error of the travel trajectory model and the error of the target model. Decide whether to start estimating the target position. In particular, in this embodiment, estimation of the target position is started by using the calculation result output from the model with the smaller error of the traveling locus model and the target model as the other input of the traveling locus model and the target model. To do.

  For example, the model error determination unit 18 compares the error of the travel locus model with the error of the target model, and when the error of the travel locus model is smaller than the error of the target model, the calculation output from the travel locus model The estimation of the target position is started using the result as an input of the target model. Further, when the error of the target model is smaller than the error of the traveling locus model, estimation of the target position is started using the calculation result output from the target model as an input of the traveling locus model.

  When the correction of the target position is completed, the matching unit 19 matches the target position corrected by the target position correcting unit 17 with the map information, and corrects the trajectory and the target until the matching degree becomes a predetermined value or more. Control is performed so as to repeatedly correct the position of the mark. At this time, not only the corrected target position but also the corrected trajectory may be matched with the map information.

  Specifically, the matching unit 19 superimposes the position of the corrected target with the map information, collates, and determines whether or not the target position matches the position of the target recorded in the map information by map matching. To do. For example, a difference between the corrected coordinates of the position of the white line and the coordinates of the position of the white line recorded in the map information is obtained, and the determination is made based on whether or not the difference is a predetermined value or less. When the matching unit 19 determines that the corrected target position does not match the target position recorded in the map information and the matching degree is less than a predetermined value, the trajectory correction is performed. And control so that the correction of the position of the target is repeatedly performed. On the other hand, when the corrected position of the target and the position of the target recorded in the map information match and it is determined that the matching degree is equal to or greater than a predetermined value, the matching with the map information is terminated. When the matching is completed, the matching unit 19 links to the map database 6 and records the corrected target position in the map database 6.

  The self-position estimating unit 20 estimates the self-position using the position of the target recorded in the map database 6 by the matching unit 19. In this embodiment, since the position of the target on the map is estimated with high accuracy, the self-position estimation unit 20 can estimate the self-position with high accuracy by using the distance and direction from the host vehicle to the target. it can. When the vehicle's own position is estimated using the position of the target, when the vehicle travels around the target, the distance and direction to the target are detected by a sensor, the detected distance and direction, and This means that the position of the host vehicle on the map is estimated from the acquired position of the target.

  The communication device 7 is a device for communicating with the server 25.

[Trajectory correction]
Next, a trajectory correction method will be described with reference to FIGS. FIG. 4 is a flowchart showing trajectory correction processing.

  As shown in FIG. 4, in step S1, the trajectory calculation unit 13 acquires the distance L to the target, the direction φ of the target, and the odometry.

  In step S2, the trajectory calculation unit 13 initializes the vehicle orientation θ and the control parameter q with predetermined values (for example, 0).

  In step S <b> 3, the trajectory calculation unit 13 inputs the initialized vehicle orientation θ and the control parameter q to the travel locus model X, and calculates a trajectory. A trajectory 21 shown in FIG. 5 is an initial trajectory. Further, the trajectory calculation unit 13 may use the target data output from the target model as an input of the travel locus model.

  In FIG. 5, coordinates indicated by dotted lines are map coordinate systems, and coordinates indicated by solid lines are vehicle coordinate systems. The y-axis in the vehicle coordinate system indicates the traveling direction of the host vehicle, and the x-axis indicates the lateral direction orthogonal to the traveling direction. Coordinates (x1, y1) to coordinates (x5, y5) shown in FIG. 5 are self-positions on the trajectory 21. The vehicle orientations θ1 to θ5 shown in FIG. 5 indicate yaw angles from the map coordinate system. The angles Φ1 to Φ5 are angles indicating the direction of the target 30 with respect to the traveling direction of the host vehicle as viewed from the vehicle coordinate system. The distances L1 to L5 are distances from the host vehicle to the target 30 as viewed from the vehicle coordinate system. In the present embodiment, the description has been given using a predetermined coordinate system. However, the present invention is not limited to the map coordinate system or the vehicle coordinate system, and other coordinate systems may be used, or one coordinate system may be used. It may be used.

  As shown in FIG. 5, the host vehicle uses a camera 2 or a laser range finder 3 while traveling in a section of coordinates (x1, y1) to coordinates (x5, y5). ~ L5 is measured. The own vehicle detects the orientations θ1 to θ5 of the vehicle using the gyro sensor 5 at the same time as the time when the distances L1 to L5 are measured. In the example illustrated in FIG. 5, a state in which the distances from five different travel points to the target 30 and the direction of the target 30 are acquired is shown, but it is not necessary to limit to five points. The angles Φ1 to Φ5 can also be obtained using the gyro sensor 5.

  In step S4, the target position calculation unit 14 arranges the target data on the map based on the trajectory data calculated in step S3, and calculates the position of the target. At this time, target data output by the target model may be used as the target data. On the trajectory 21, if the self position at the i-th time is Xi, the self position Xi is expressed as follows using equation (5).

The position of the target 30 _(Cx i, Cy _i) When the position of the target 30 _(Cx i, Cy _i), using equation (6) is expressed as follows.

(Lx _i , Ly _i ) in Equation (6) is the distance from the host vehicle to the target 30 at the i-th time, and includes an element of the angle Φ. As shown in Expression (6), the target position calculation unit 14 rotates the distance (Lx _i , Ly _i ) from the host vehicle to the target 30 by the vehicle orientation θ _i . Then, the target position calculation unit 14 adds the distance to the coordinates (x _i , y _i ) on the trajectory 21, translates the coordinates (x _i , y _i ), and converts them into a map coordinate system. The position of the target 30 is calculated for each point.

  In step S5, the trajectory correction unit 15 obtains a difference between the target position calculated in step 4 and the target position recorded in the map information. That is, in step S4, the difference between the target data arranged on the basis of the trajectory data and the map information is calculated. In addition, you may obtain | require dispersion | distribution of the position of a target instead of a difference. Further, the difference calculated here is estimated as an error of the travel locus model.

  In step S6, the trajectory correction unit 15 determines whether or not the difference calculated in step S5 is equal to or less than a predetermined value. If the difference is less than or equal to the predetermined value, the trajectory correction process ends and trajectory data is output. On the other hand, if the difference is greater than the predetermined value, the process proceeds to step S7.

  In step S7, the trajectory correction unit 15 adjusts the trajectory data with a predetermined model so that the difference becomes small, and repeatedly adjusts the trajectory data until the difference becomes smaller than a predetermined value. That is, trajectory correction is repeatedly performed so that the position of the target calculated in step S4 matches the position of the target recorded in the map information. For example, the trajectory 21 is corrected by using an optimization method such as the Gauss-Newton method, gradually adjusting the vehicle orientation θ and the control parameter q to gradually approach the true value in the repeated calculation. As a result, the trajectory 21 shown in FIG. 5 is corrected as shown in FIG. In FIG. 5, the trajectory 21 is calculated so as to be shifted left and right, but in FIG. 6, the trajectory 21 is corrected to a smooth line.

  Here, since the control parameter q is correlated with the control parameter p when adjusting the control parameter q, the value of the control parameter q is set and adjusted based on the value of the control parameter p. Thereby, the time required for the process of correcting the trajectory can be shortened. Furthermore, since the range for adjusting the control parameter q can be narrowed, robustness can be improved.

[Target position correction]
Next, a target position correction method will be described with reference to FIGS. FIG. 7 is a flowchart showing target position correction processing.

  As shown in FIG. 7, in step S <b> 10, the target calculation unit 16 acquires the distance L to the target obtained by the sensor of the host vehicle, the direction φ of the target, and the trajectory. The trajectory may be one corrected by the trajectory correction unit 15 or another trajectory.

  In step S20, the target calculation unit 16 initializes the vehicle orientation θ and the control parameter p with predetermined values (for example, 0).

  In step S <b> 30, the target calculation unit 16 calculates the target by inputting the initialized vehicle orientation θ and the control parameter p into the target model L. In FIG. 8, the white line 40 is shown as an example of the target, and the white line 40 is an initial white line. Further, the target calculation unit 16 may use the trajectory data output from the travel locus model as an input of the target model L.

  As shown in FIG. 8, the host vehicle uses a camera 2 or a laser range finder 3 while traveling in a section of coordinates (x1, y1) to coordinates (x5, y5), and distances L1 to L1 from the host vehicle to the white line 40. L5 is measured. The own vehicle detects the orientations θ1 to θ5 of the vehicle using the gyro sensor 5 at the same time as the time when the distances L1 to L5 are measured. In the example illustrated in FIG. 8, a state in which the distance from the five different traveling points to the white line 40 and the direction of the white line 40 are obtained is shown, but it is not necessary to limit to five points. The angles Φ1 to Φ5 can also be obtained using the gyro sensor 5.

  In step S40, the target calculation unit 16 calculates the position of the target by arranging the target data on the map with reference to the travel locus data obtained by the odometry of the host vehicle. At this time, trajectory data output by the travel locus model may be used as the travel locus data. On the white line 40, assuming that the position of the white line at the i-th time is Li, the position Li of the white line is expressed as follows using equation (7).

When the position of the white line 40 is (Cx _i , Cy _i ), the position (Cx _i , Cy _i ) of the white line 40 is expressed as follows using the equation (8).

(Lx _i , Ly _i ) in Expression (8) is the distance from the host vehicle to the white line 40 at the i-th time, and includes an element of the angle Φ. As shown in Expression (8), the target calculation unit 16 rotates the distance (Lx _i , Ly _i ) from the host vehicle to the white line 40 by the vehicle orientation θ _i . Then, the target calculation unit 16 adds the distance to the coordinates (x _i , y _i ) on the trajectory 21, translates the coordinates (x _i , y _i ), and converts them into a map coordinate system. The position of the white line 40 is calculated every time.

  In step S50, the target position correcting unit 17 obtains a difference between the position of the target calculated in step S40 and the position of the target recorded in the map information. That is, the difference between the target data arranged on the basis of the travel locus data and the map information is calculated in step S40. In addition, you may obtain | require dispersion | distribution of the position of a target instead of a difference. Further, the difference calculated here is estimated as an error of the target model.

  In step S60, the target position correcting unit 17 determines whether or not the difference calculated in step S50 is equal to or less than a predetermined value. If the difference is less than or equal to the predetermined value, the target correction process is terminated and the target data is output. On the other hand, if the difference is greater than the predetermined value, the process proceeds to step S70.

  In step S70, the target position correcting unit 17 adjusts the target data with a predetermined model so that the difference becomes small, and repeatedly adjusts the target data until the difference becomes smaller than a predetermined value. That is, the correction of the target is repeatedly performed so that the position of the target calculated in step S40 matches the position of the target recorded in the map information. For example, the white line 40 is corrected by using an optimization method such as the Gauss-Newton method and gradually adjusting the vehicle orientation θ and the control parameter p to gradually approach the true value in the repeated calculation. As a result, the white line shown in FIG. 8 is corrected as shown in FIG. In FIG. 8, the white line 40 is calculated so as to be shifted left and right, but in FIG. 9, the white line 40 is corrected to a smooth line.

  Here, since the control parameter p is correlated with the control parameter q when the control parameter p is adjusted, the value of the control parameter p is set and adjusted based on the value of the control parameter q. Thereby, the time required for the process of correcting the position of the target can be shortened. Furthermore, since the range for adjusting the control parameter p can be narrowed, robustness can be improved.

[Self-position estimation process]
Next, with reference to the flowchart of FIG. 10, a processing procedure of the self-position estimation process by the self-position estimation apparatus 1 according to the present embodiment will be described.

In step S101, the camera 2 and laser range finder 3 obtains the distance L _i and orientation theta _i of the vehicle from a different travel points as shown in FIG. 5 to the target 30. In the present embodiment, i indicates time and its value is 1-5.

  In step S102, the trajectory calculation unit 13 initializes the vehicle orientation θ and the control parameter q with predetermined values.

  In step S103, the trajectory calculation unit 13 inputs the initial value set in step S102 to the travel locus model X, and calculates the initial trajectory 21 as shown in FIG. Then, the target position calculation unit 14 uses the calculated trajectory 21 to calculate the position of the target 30 on the map. That is, the target data is arranged on the map on the basis of the trajectory data, and the position of the target 30 is calculated. When the position of the target is calculated, the trajectory correction unit 15 obtains a difference between the calculated position of the target and the position of the target recorded in the map information, and uses this difference as an error in the travel locus model. calculate.

  In step S104, the target calculation unit 16 calculates the target data around the host vehicle by inputting the target data obtained by the sensor of the host vehicle into the target model L. Then, the calculated target data is arranged on a map on the basis of the travel locus data obtained by odometry of the host vehicle, and the position of the target is calculated. For example, as shown in FIG. 8, the white line 40 is arranged on the map with the trajectory 21 as a reference, and the position of the white line 40 is calculated. When the position of the target is calculated, the target position correcting unit 17 obtains a difference between the calculated position of the target and the position of the target recorded in the map information, and uses this difference of the target model. Calculate as error. In step S103, the target 30 is described as an example. In step S104, the white line 40 is described as an example. However, this is only an example, and steps S103 and 104 calculate an error for different targets. Instead of calculating the error for the same target.

  In step S105, the model error determination unit 18 compares the error of the travel locus model calculated in step S103 with the error of the target model calculated in step S104. It is determined whether it is smaller than the error. As a result of the determination, if the error of the target model is smaller than the error of the travel locus model, the process proceeds to step S106, and estimation of the target position is started using the calculation result output from the target model as the input of the travel locus model. To do. On the other hand, if the error of the target model is greater than or equal to the error of the travel locus model, the process proceeds to step S110, and the estimation of the target position is started using the calculation result output from the travel locus model as the input of the target model. .

  In step S <b> 106, the trajectory calculation unit 13 calculates a trajectory using the travel locus model X. Specifically, the trajectory calculation unit 13 calculates a trajectory using the target data output from the target model as an input of the travel locus model. For example, the trajectory is calculated by inputting the position of the target calculated in step S104 or step S111 described later to the travel locus model. In this case, the trajectory calculation unit 13 calculates the trajectory by calculating backward from the position of the target. Further, the trajectory calculation unit 13 may input the initial value set in step S102 to the travel locus model X and calculate the initial trajectory 21 as shown in FIG. Further, the trajectory data may be calculated based on the travel locus data obtained by the odometry of the host vehicle.

In step S107, the target position calculation unit 14 calculates the position of the target on the map using the trajectory calculated in step S106. First, the target position calculation unit 14 acquires a distance L _i from each point on the trajectory 21 to the target 30 and acquires a vehicle orientation θ _i when the distance is acquired. Then, the target position calculation unit 14 calculates the position of the target 30 for each point using the distance L _i to the target 30 and the direction θ _i of the vehicle. That is, the target data is arranged on the map on the basis of the trajectory data, and the position of the target 30 is calculated. At this time, target data output by a target model described later may be used as target data to be arranged on the map.

  In step S108, the trajectory correction unit 15 determines whether or not the position of the target 30 calculated in step S107 matches the position of the target recorded in the map information. As a determination method, a difference between the calculated position of the target 30 and the position of the target recorded in the map information may be obtained, and the determination may be made based on whether the difference is equal to or less than a predetermined threshold value. Alternatively, the variance of the calculated position of the target 30 may be calculated, and determination may be made based on whether or not the variance is equal to or less than a predetermined threshold. If the trajectory correction unit 15 determines that the calculated position of the target 30 does not match the position of the target recorded in the map information, the process proceeds to step S109. On the other hand, when it is determined that the calculated position of the target 30 matches the position of the target recorded in the map information, the trajectory correction is terminated, and the process proceeds to step S110.

  In step S109, the trajectory correction unit 15 corrects the trajectory 21 by adjusting the vehicle orientation θ and the control parameter q. Thereafter, in steps S106 to S109, the trajectory correction unit 15 uses an optimization method such as the Gauss-Newton method to gradually approach the true value in the repeated calculation while adjusting the vehicle orientation θ and the control parameter q. The trajectory is corrected using a method that goes on. For example, the trajectory data is adjusted by a predetermined model so that the difference between the position of the target 30 and the position of the target recorded in the map information becomes small.

  In step S110, the target calculation unit 16 calculates target data using the target model L. Specifically, the target calculation unit 16 calculates target data using the calculation result output from the travel locus model as an input of the target model. For example, the target data output by the travel locus model in steps S103 and 107 is used as the input of the target model L to calculate target data. The target calculation unit 16 may calculate target data by inputting the vehicle orientation θ and the control parameter p to the target model L. Furthermore, the target calculation unit 16 may input target data obtained by a sensor of the host vehicle to the target model L to calculate target data around the host vehicle.

  In step S111, the target calculation unit 16 calculates the position of the target by arranging the target data on the map with reference to the travel locus data obtained by the odometry of the host vehicle. At this time, the trajectory data output from the travel trajectory model in steps S103 and S106 may be used as the travel trajectory data. For example, as shown in FIG. 8, the white line 40 may be arranged on the map with the trajectory 21 as a reference, and the position of the white line 40 may be calculated.

  In step S112, the target position correcting unit 17 determines whether or not the position of the white line 40 calculated in step S111 matches the position of the white line recorded in the map information. As a determination method, a difference between the calculated position of the white line 40 and the position of the white line recorded in the map information is obtained, and the determination is made based on whether the difference is equal to or less than a predetermined value. Alternatively, the variance of the calculated position of the white line 40 may be calculated, and determination may be made based on whether the variance is equal to or less than a predetermined threshold. If the target position correction unit 17 determines that the calculated position of the white line 40 does not match the position of the white line recorded in the map information, the process proceeds to step S113. On the other hand, when it is determined that the calculated position of the white line 40 matches the position of the white line recorded in the map information, the correction of the position of the white line is finished and the process proceeds to step S114.

  In step S113, the target position correction unit 17 corrects the white line 40 by adjusting the control parameter p. Thereafter, in step S110 to step S113, the target position correcting unit 17 uses an optimization method such as the Gauss-Newton method and gradually approaches the true value in the repeated calculation while adjusting the control parameter p. The position of the white line 40 is corrected using a method. For example, the target data is adjusted by a predetermined model so that the difference between the position of the white line 40 and the position of the white line recorded in the map information becomes small.

  In step S114, the matching unit 19 matches the corrected position of the white line with the map information, and determines whether or not the matching degree is a predetermined value or more. Specifically, the matching unit 19 overlaps the position of the corrected white line with the map information and collates it to determine whether or not it matches the white line recorded in the map information. If it is determined that the corrected white line does not match the map information and the degree of matching is less than a predetermined value, the process returns to step S106, and trajectory correction and white line position correction are repeated. When the process returns from step S114 to step S106, a trajectory is calculated based on the corrected target position, and the target position is calculated. On the other hand, if the corrected white line matches the map information and it is determined that the matching degree is greater than or equal to a predetermined value, the matching with the map information is terminated and the process proceeds to step S115.

  As described above, in the present embodiment, the calculation result output by the travel locus model in steps S106 and 107 is used as the input of the target model in steps S110 and 111. Moreover, the calculation result output by the target model of steps S110 and 111 is used as the input of the travel locus model of steps S106 and 107. Therefore, in the self-position estimation apparatus 1 according to the present embodiment, the calculation result output from one of the travel locus model and the target model is used as the other input of the travel locus model and the target model.

  In step S115, the matching unit 19 links to the map database 6 and records the corrected position of the white line in the map database 6. If the position of the white line is corrected and the position of other road structures or targets needs to be corrected, those positions are also corrected and recorded.

  In step S116, the self-position estimation unit 20 estimates the self-position using the position of the white line 40 recorded in the map database 6 in step S115, and ends the self-position estimation process according to the present embodiment. Since the position of the white line 40 on the map is accurately estimated in the self-position estimation process of the present embodiment, the self-position estimation unit 20 uses the distance and direction from the host vehicle to the white line 40 to determine the self-position estimation unit 20. Can be estimated with high accuracy.

[Effect of the first embodiment]
As described above in detail, the self-position estimation apparatus 1 according to the present embodiment calculates the error of the travel locus model and the error of the target model. The target position is estimated using either the calculation result output from the travel trajectory model or the calculation result output from the target model, depending on the error of the travel trajectory model and the error of the target model. Decide to start. Thereby, since the estimation of the target position can be started in consideration of the error of the travel locus model and the error of the target model, the self position of the host vehicle can be estimated with high accuracy.

  For example, as shown in FIG. 11, there may be a positional shift of several centimeters, for example, about 10 cm between the white line 91 on the actual road and the white line 93 recorded in the map information. Therefore, if the map information is applied as it is in the actual road environment to estimate the position of the white line and calculate the self-position of the vehicle, an error occurs between the self-position and the actual road. In addition, depending on the map, the target information may be insufficient, the relative positional relationship with the target such as a white line may not be known, and the position of the host vehicle may not be accurately estimated.

  However, in the self-position estimation apparatus 1 according to the present embodiment, estimation of the target position is started in consideration of the error of the travel locus model and the error of the target model. Even if it is insufficient, the self position of the host vehicle can be accurately estimated.

  In the self-position estimation apparatus 1 according to the present embodiment, the calculation result output from the model with the smaller error of the traveling locus model and the target model is used as the other input of the traveling locus model and the target model. Start estimating the target position. As a result, estimation of the target position can be started based on the calculation result output from the model with the smaller error, so that the self-position of the host vehicle can be estimated accurately and the self-position can be estimated earlier. It becomes.

  Furthermore, since the self-position can be accurately estimated, for example, when automatic driving control or driving support control is executed, the position of the vehicle can be accurately controlled, so that appropriate position control that matches the actual target can be executed. It becomes like this. In addition, since it is possible to execute appropriate driving support that matches the actual target, it is possible to execute automatic driving control or driving support control that suppresses a sense of discomfort given to the occupant.

[Second Embodiment]
Hereinafter, a second embodiment of the present invention will be described with reference to the drawings. In the description of the drawings, the same parts as those in the first embodiment are denoted by the same reference numerals, and the description thereof is omitted.

[Configuration of self-position estimation device]
In the self-position estimation apparatus according to the present embodiment, only the processing performed by the model error determination unit 18 is different from that of the first embodiment, and other configurations of the self-position estimation apparatus are shown in FIG. Is the same.

  The model error determination unit 18 uses either the calculation result output from the travel trajectory model or the calculation result output from the target model according to the magnitude of the error of the travel trajectory model and the error of the target model. Decide whether to start estimating the target position. In particular, in the present embodiment, predetermined thresholds are set for the error of the travel locus model and the error of the target model, respectively. Then, depending on whether the error of the travel trajectory model and the error of the target model are larger than a predetermined threshold, the calculation result output from the travel trajectory model or the calculation result output from the target model is used. Determine whether to start estimating the target position.

  The threshold value to be set is a threshold value for determining whether or not correction of the calculated target position is unnecessary, and may be set through experiments or simulations, or may be set arbitrarily. Good.

  Here, with reference to FIG. 12, a method for determining which of the calculation result of the travel locus model and the calculation result of the target model is used to start the estimation of the target position will be described. As shown in FIG. 12A, the model error determination unit 18 corrects the calculated position of the target when the error of the travel locus model and the error of the target model are both below a predetermined threshold. It is determined that it is unnecessary.

  As shown in FIG. 12B, the model error determination unit 18 uses the calculation result output from the travel locus model as the input of the target model when the error of the target model is larger than a predetermined threshold. It is determined that estimation of the target position is started. In addition, as shown in FIG. 12C, when the error of the travel trajectory model is larger than a predetermined threshold, estimation of the target position is started using the calculation result output from the target model as the input of the travel trajectory model. Judge that.

  Further, as shown in FIG. 12D, when both the error of the travel locus model and the error of the target model are larger than a predetermined threshold, the calculation result output from the travel locus model is input to the target model. It is determined that estimation of the target position is started. In general, the error of odometry is smaller than the displacement of the target. Therefore, if both the error in the travel trajectory model and the error in the target model are larger than the predetermined threshold, the travel trajectory model is prioritized and the target position is estimated using the calculation result output from the travel trajectory model. To start.

  If the vehicle speed of the host vehicle is less than the predetermined value, the model error determination unit 18 may start estimation of the target position using the calculation result output from the target model as an input of the travel locus model. Good. Since the trajectory error tends to increase when the vehicle speed becomes extremely low, if the vehicle speed of the host vehicle falls below the specified value, the target model is prioritized and the calculation result output from the target model The target position estimation is started using. As for the predetermined value, a vehicle speed at which a trajectory error increases may be obtained in advance through experiments and simulations, and set to that value.

  Further, not only the vehicle speed but also the trajectory error tends to increase when the road on which the vehicle is traveling is curved. Therefore, even when the host vehicle is traveling on a curve, the target model may be prioritized and estimation of the target position may be started using the calculation result output from the target model.

[Self-position estimation process]
Next, with reference to the flowchart of FIG. 13, a processing procedure of self-position estimation processing by the self-position estimation apparatus 1 according to the present embodiment will be described. However, the self-position estimation process according to the present embodiment is the same as the first embodiment except that step S105 in FIG. 10 is changed to steps S205 and 206 in FIG. is there. Therefore, detailed description of other steps is omitted.

  As shown in FIG. 13, the processing of steps S201 to S204 is executed, and the error of the travel locus model and the error of the target model are calculated. In step S205, the model error determination unit 18 determines whether correction is necessary. That is, it is determined whether at least one of the error of the travel locus model calculated in step S203 and the error of the target model calculated in step S204 is larger than a predetermined threshold value.

  As shown in FIG. 12A, the model error determination unit 18 determines that the correction is unnecessary when both the error of the travel locus model and the error of the target model are equal to or less than a predetermined threshold value. Proceed to step S215. On the other hand, if one or both of the error of the travel locus model and the target model is larger than the predetermined threshold, that is, in the case of B, C, and D in FIG. 12, it is determined that correction is necessary. The process proceeds to step S206.

  In step S206, the model error determination unit 18 determines whether or not to start the estimation of the self position using the calculation result output from the target model. When the error of the travel locus model is larger than the predetermined threshold, that is, in the case of C in FIG. 12, the model error determination unit 18 starts estimating the target position using the calculation result output from the target model. Determine and proceed to step S207.

  On the other hand, when the error of the target model is larger than the predetermined threshold, that is, in the case of B in FIG. 12, the model error determination unit 18 estimates the target position using the calculation result output from the traveling locus model. It determines with starting, and progresses to step S211. Further, even when the error of the travel locus model and the error of the target model are both larger than a predetermined threshold, that is, in the case of D in FIG. 12, the target position is estimated using the calculation result output from the travel locus model. The process proceeds to step S211. Thereafter, when the processes of steps S207 to 217 are executed and the self-position is estimated, the self-position estimation process according to the present embodiment is ended.

[Effects of Second Embodiment]
As described above in detail, in the self-position estimation apparatus 1 according to the present embodiment, predetermined thresholds are set for the error of the travel locus model and the error of the target model, respectively. Then, depending on whether the error of the travel trajectory model and the error of the target model are larger than a predetermined threshold, the calculation result output from the travel trajectory model or the calculation result output from the target model is used. Determine whether to start estimating the target position. Thereby, since the estimation of the target position can be started in consideration of the error of the travel locus model and the error of the target model, the self position of the host vehicle can be estimated with high accuracy.

  Further, in the self-position estimation apparatus 1 according to the present embodiment, when the error of the target model is larger than a predetermined threshold, the calculation result output from the travel locus model is used as the target model input as the target position. Start estimation. Thereby, even when the error of the target model is large, the estimation of the target position can be started using the calculation result output from the traveling locus model, so that the self position of the host vehicle can be estimated with high accuracy.

  Furthermore, in the self-position estimation apparatus 1 according to the present embodiment, when the error of the travel locus model is larger than a predetermined threshold, the calculation result output from the target model is used as the input of the travel locus model as the target position. Start estimation. Thereby, even when the error of the travel locus model is large, the estimation of the target position can be started using the calculation result output from the target model, so that the self position of the host vehicle can be estimated with high accuracy.

  Further, in the self-position estimation apparatus 1 according to the present embodiment, when the error of the travel locus model and the error of the target model are larger than a predetermined threshold, the calculation result output from the travel locus model is displayed on the target model. Start estimation of target position as input. As a result, when the error between both the travel trajectory model and the target model is large, the estimation of the target position can be started using the calculation result output from the travel trajectory model considered to have a smaller error. It is possible to accurately estimate the self-position of the vehicle.

  Moreover, in the self-position estimation apparatus 1 according to the present embodiment, when the vehicle speed of the host vehicle is less than a predetermined value, the calculation result output from the target model is used as the input of the travel locus model to estimate the target position. To start. This makes it possible to start estimating the target position using the calculation result output from the target model when the vehicle speed of the host vehicle is low and there is a high possibility that the trajectory error will increase. Can be estimated with high accuracy.

  It is also possible to record the corrected target and use it for the next or subsequent driving support. As a result, the travel support can be executed using the corrected target data, so that the appropriate travel support suitable for the actual target can be executed when the next or subsequent travel support is executed. In addition, by recording the corrected target data, it is not necessary to estimate the target when traveling on the same road, so that the calculation load can be reduced.

  The above-described embodiment is an example of the present invention. For this reason, the present invention is not limited to the above-described embodiment, and even if it is a form other than this embodiment, as long as it does not depart from the technical idea of the present invention, it depends on the design. Of course, various modifications are possible.

DESCRIPTION OF SYMBOLS 1 Self-position estimation apparatus 2 Camera 3 Laser range finder 4 GPS receiver 5 Gyro sensor 6 Map database 7 Communication apparatus 10 Controller 11 Traveling track model creation part 12 Target model creation part 13 Trajectory calculation part 14 Target position calculation part 15 Trajectory Correction unit 16 Target calculation unit 17 Target position correction unit 18 Model error determination unit 19 Matching unit 20 Self-position estimation unit 25 Server 30 Target 40 White line

Claims (8)

A travel locus model obtained by inputting travel locus data obtained by odometry of the own vehicle, adjusting the travel locus data obtained by the odometry by a predetermined model, and outputting the adjusted result as trajectory data of the own vehicle;
The target data obtained by the sensor of the host vehicle is input, the target data obtained by the sensor is adjusted by a predetermined model, and the adjusted result is output as the target data around the host vehicle. In a self-position estimation method of a self-position estimation device that estimates a target position around the host vehicle using a model and estimates the self-position of the host vehicle by comparing the target position and map information,
Calculate the error of the travel locus model and the error of the target model,
Depending on the error of the travel locus model and the error of the target model, either the calculation result output from the travel locus model or the calculation result output from the target model is used. A self-position estimation method characterized by determining whether to start position estimation.   Starting estimation of the target position using the calculation result output from the model with the smaller error of the traveling locus model and the target model as the other input of the traveling locus model and the target model. The self-position estimation method according to claim 1. A predetermined threshold is set for each of the error of the travel locus model and the error of the target model,
Depending on whether the error of the travel locus model and the error of the target model are larger than the predetermined threshold, either the calculation result output from the travel locus model or the calculation result output from the target model is used. The method of claim 1, further comprising determining whether to start estimating the target position.   When the error of the target model is larger than the predetermined threshold value, estimation of the target position is started by using the calculation result output from the traveling locus model as an input of the target model. The self-position estimation method according to claim 3.   When the error of the travel locus model is larger than the predetermined threshold, the estimation of the target position is started using the calculation result output from the target model as an input of the travel locus model. The self-position estimation method according to claim 3 or 4.   When the error of the travel locus model and the error of the target model are larger than the predetermined threshold, estimation of the target position is started using the calculation result output from the travel locus model as an input of the target model. The self-position estimation method according to any one of claims 3 to 5, wherein:   The target position estimation is started by using the calculation result output from the target model as an input of the travel locus model when the vehicle speed of the host vehicle is less than a predetermined value. 7. The self-position estimation method according to any one of 6 above. A travel locus model obtained by inputting travel locus data obtained by odometry of the own vehicle, adjusting the travel locus data obtained by the odometry by a predetermined model, and outputting the adjusted result as trajectory data of the own vehicle;
The target data obtained by the sensor of the host vehicle is input, the target data obtained by the sensor is adjusted by a predetermined model, and the adjusted result is output as the target data around the host vehicle. In a self-position estimation device that estimates a target position around a host vehicle using a model and estimates the self-position of the host vehicle by comparing the target position with map information,
The self-position estimation device
Calculate the error of the travel locus model and the error of the target model,
Depending on the error of the travel locus model and the error of the target model, either the calculation result output from the travel locus model or the calculation result output from the target model is used. A self-position estimation apparatus that determines whether to start position estimation.