JP6645936B2 - State estimation device - Google Patents

Description

  The present disclosure relates to a state estimation device mounted on a vehicle and estimating a position of the vehicle.

2. Description of the Related Art In driving assistance and automatic driving in which a vehicle assists at least a part of a driving operation performed by a driver of a vehicle, there is known a technology related to a lane recognition that recognizes a lane on which the vehicle is traveling.
For example, Patent Document 1 discloses the following technology. That is, a feature such as a sign or a white line detected by the on-board camera is detected as a detected feature in each observation cycle set in advance. The position of the own vehicle in the current observation cycle is predicted from the state of the own vehicle such as the estimated position and the speed vector estimated in the previous observation cycle, and is set as a predicted position. Based on the predicted position of the vehicle, feature data registered in the map database and estimated to be the same as the detected feature is associated with the detected feature. By correcting the predicted position so that the relative positional relationship between the detected feature and the feature data associated with the predicted position of the own vehicle matches each other, the own vehicle in the current observation cycle is corrected. Is estimated.

JP 2008-249639 A

  However, in the device described in Patent Literature 1, when the actual position of the own vehicle, particularly the vertical position along the traveling direction of the own vehicle and the predicted position are significantly different, the road shape specified from the white line or the like causes An error occurs in the correspondence between the relative position relationship between the position feature and the vehicle detected by the relationship detection unit and the position represented by the feature data associated with the position feature. As a result, when the state of the host vehicle is estimated based on an erroneous correspondence, an error is likely to occur.

  The present disclosure provides a technique for suppressing a deviation of a lateral position and a vehicle direction of a vehicle even when a vertical position is corrected.

  The state estimation device (1) of the present disclosure repeatedly estimates a vehicle state, which is mounted on a vehicle and includes a position and a traveling direction of the vehicle, for each observation cycle having a preset cycle, and a state prediction unit (S110). Map acquisition unit (S120), feature detection unit (S140, S200), relationship detection unit (S150), association unit (S160, S230), correction calculation unit (S210), first estimation unit (S170), and second Estimation units (S240, S250) are provided.

  The state prediction unit generates a predicted state that is a vehicle state in the current observation cycle, which is predicted using the estimated state that is the vehicle state estimated in the previous observation cycle. The map acquisition unit acquires map data including feature data representing a feature existing around the predicted position and shape data representing a road shape, according to the predicted position that is the position of the vehicle included in the predicted state. The feature detector is a feature that defines a position feature and a road shape that are at least the installation location registered in the feature data from the detection result of the optical sensor (5) for each observation cycle. Detects shaped features. The relation detecting unit detects a relative positional relation between the position feature and the vehicle. The associating unit, assuming that the vehicle exists at the predicted position, the position feature and the feature estimated to be the same based on the relative positional relationship between the position feature and the vehicle detected by the relationship detecting unit. Make correspondence with data. The correction calculation unit calculates, as a correction amount, a difference between the relative position of the position feature with respect to the vehicle and the relative position of the position represented by the feature data associated with the position feature by the association unit with the predicted position. The first estimating unit sets the position along the traveling direction of the vehicle as a vertical position, and represents the relative position of the position feature with respect to the vehicle, and the feature data associated with the position feature by the associating unit for the predicted position. By using the difference between the position and the relative position as a correction amount and correcting the predicted position using the correction amount, at least the vertical position of the vehicle in the current measurement cycle is estimated. The second estimating unit, when the vertical correction amount that is the vertical position correction amount among the correction amounts calculated by the correction calculation unit is equal to or less than a preset correction amount threshold, sets the predicted position as the comparison position; When the vertical correction amount is larger than the correction amount threshold value, the point where the vertical position of the predicted position is corrected by the vertical correction amount is set as a comparison position, and the road shape represented by the shape data at the comparison position acquired by the map acquisition unit. And, by alignment with the road shape indicated by the shape feature detected by the feature detection unit, at least one of a lateral position that is a position along the vehicle width direction and a vehicle direction that is the traveling direction of the vehicle. Make an estimate. The state estimating device sets the estimation results of the first estimating unit and the second estimating unit as the estimated state in the current observation cycle.

  In such a configuration, when the vertical correction amount is larger than the correction amount threshold, the position of the feature indicated by the map data in the corrected vertical position of the vehicle and the position of the feature obtained by the optical sensor are compared. The lateral position and the vehicle direction of the vehicle are corrected based on. Thus, even when the vertical position is corrected, it is possible to suppress the lateral position and the vehicle direction of the vehicle from shifting.

  It should be noted that reference numerals in parentheses described in this column and in the claims indicate a correspondence relationship with specific means described in the embodiment described below as one aspect, and the technical scope of the present disclosure will be described. It is not limited.

It is a block diagram showing composition of a state estimating device and peripheral equipment. FIG. 3 is a diagram illustrating a camera coordinate system. It is a figure showing an image coordinate system. It is a figure showing the image of the image picturized by the stereo camera. It is a flowchart which shows a state estimation process. It is a figure showing the state where a road boundary cannot be detected by an optical sensor. It is a flowchart which shows a center position detection process. FIG. 9 is a diagram illustrating a state in which a sign is deviated from an angle of view in detection of a terrestrial object by an optical sensor and interpolation is required. It is a figure showing the outline of the method of associating in a prediction position. It is a figure showing the outline of the re-association method in a comparison position. FIG. 14 is a diagram illustrating an example of an image captured by an optical sensor according to a modification.

Hereinafter, embodiments of the present disclosure will be described with reference to the drawings.
[1. First Embodiment]
[1-1. Constitution]
The driving support system 2 of the present embodiment is mounted on a vehicle and includes a state estimating device 1 as shown in FIG. The driving support system 2 may further include a GNSS receiver 3, a map memory 4, an optical sensor 5, a state quantity sensor 6, and a control device 7. Hereinafter, a vehicle equipped with the state estimation device 1 is referred to as a host vehicle.

  The GNSS receiver 3 is a device that receives a transmission radio wave from a GNSS artificial satellite via an antenna. The GNSS receiver 3 measures the absolute position of the own vehicle and outputs the measured absolute position to the state estimating device 1 as position data.

  The map memory 4 stores map data of a road on which the vehicle can travel for use in normal navigation. The map data includes feature data representing a feature and shape data representing a road shape. Here, the feature data includes sign data representing a road sign installed along a road on which the vehicle can travel. The sign data includes the location of the road sign on the map, the type of sign, the height of the sign with respect to the road on which the vehicle is traveling, the sign size, the characters displayed on the sign, and the data of the pillars supporting the sign. including. The shape data includes boundary data that is data of a position of a boundary display indicating a boundary of a road on which the own vehicle is traveling or a boundary of a lane.

  The optical sensor 5 acquires an image including a feature in front of the own vehicle. The feature here includes a location feature such as a road sign ahead of the own vehicle and a shape feature such as a road boundary and a lane boundary. The optical sensor 5 outputs an image representing a feature in front of the own vehicle to the state estimation device 1. In the present embodiment, the optical sensor 5 is a stereo camera which is a pair of cameras provided on the same horizontal line and arranged so that their optical axes are parallel. It is assumed that the pair of cameras constituting the stereo camera have the same focal length. Hereinafter, as shown in FIG. 2, the coordinate system in the real space based on the position where the stereo camera of the vehicle is installed is referred to as a camera coordinate system, and the coordinates in the camera coordinate system are represented by (Xc, Yc, Zc). Here, Xc represents the horizontal coordinates captured by the stereo camera, Yc represents the vertical coordinates captured by the stereo camera, and Zc represents the optical axis direction captured by the stereo camera. Here, when Xc is oriented in the optical axis direction of the stereo camera, the coordinate value increases as going to the left, and as for Yc, the coordinate value increases as going to the upper direction. On the other hand, the coordinate system in the captured image is an image coordinate system as shown in FIG. 3, and the coordinates in the image coordinate system are represented by (u, v). The image center is (cu, cv). Here, u represents the coordinates of the captured image in the horizontal direction, and the coordinate value increases toward the right of the captured image. v represents the coordinates in the vertical direction of the captured image, and the value of the coordinates increases toward the lower side of the captured image. Further, an image as shown in FIG. 4 can be acquired by a stereo camera.

  The state quantity sensor 6 detects the state of the own vehicle such as the vehicle speed and the yaw rate of the own vehicle. The state quantity sensor 6 outputs a detected vehicle state quantity representing the detected state of the vehicle to the ECU. The state quantity of the own vehicle includes the speed and acceleration of the own vehicle, and the state quantity sensor 6 is, specifically, a speed sensor and an acceleration sensor.

The control device 7 is a device that realizes functions such as automatic driving and driving support by controlling the accelerator, brake, and steering of the own vehicle according to the input from the state estimation device 1.
The state estimation device 1 estimates the position of the own vehicle based on the input map data, the image, and the amount of the own vehicle state, and outputs the estimated position of the own vehicle and the shape of the road at the position of the own vehicle to the control device 7. I do.

  The state estimating apparatus 1 is an ECU, and is mainly configured by a known microcomputer including a CPU 11 and a memory 12 that is a semiconductor memory such as a RAM, a ROM, and a flash memory. Here, the ECU is an electronic control unit of the own vehicle. Various functions of the state estimation device 1 are realized by the CPU 11 of the ECU executing a program stored in a non-transitional substantial recording medium. In this example, the memory 12 corresponds to a non-transitional substantial recording medium storing a program. When this program is executed, a method corresponding to the program is executed. Note that the number of microcomputers constituting the state estimation device 1 may be one or more.

  The technique for realizing the various functions of the state estimation device 1 is not limited to software, and some or all of the elements may be realized using one or a plurality of hardware. For example, when the above functions are implemented by electronic circuits that are hardware, the electronic circuits may be implemented by digital circuits including a large number of logic circuits, analog circuits, or a combination thereof.

[1-2. processing]
Next, a state estimating process executed by the state estimating apparatus 1 at a constant cycle will be described with reference to a flowchart of FIG. Before starting the state estimation process, the state estimation device 1 initializes the position of the own vehicle and the vehicle direction. That is, the position information acquired from the GNSS receiver 3 is set as the position of the own vehicle.

  In S110, the state estimation device 1 predicts the position and the vehicle direction of the own vehicle in the current processing cycle based on the previous estimation result and the speed and acceleration of the own vehicle acquired from the state quantity sensor 6, and the prediction is performed. Generate states as predicted states. In this case, the previous estimation result is the vertical position data corrected in S170, the horizontal position data corrected in S240, and the vehicle orientation data corrected in S250 in the previous processing cycle. is there. Further, in the case of the first processing cycle, the position and the azimuth of the own vehicle in the current processing cycle are predicted based on the initialized values. The method of estimating the position and the direction of the vehicle is a known technique, and a detailed description thereof will be omitted.

In S120, the state estimation device 1 acquires map data from the map memory 4 according to the predicted position, which is the position of the vehicle included in the predicted state generated in S110.
In S130, the state estimation device 1 acquires an image captured by the optical sensor 5 from the optical sensor 5.

  In S140, the state estimation device 1 detects a feature included in the image acquired in S130. That is, pattern matching of the shape pattern of the feature represented by the feature data stored in the memory in advance is performed on the image captured by the optical sensor 5, and the feature detected by the pattern matching is determined as the detected feature. I do. Note that the pattern matching here is a known technique, and a description thereof will be omitted.

  In S150, the state estimation device 1 detects the vertical position of the detected feature with respect to the own vehicle. Specifically, the vertical position of the detected feature with respect to the own vehicle is calculated by comparing images having parallax captured simultaneously by the stereo camera as the optical sensor 5. Here, the vertical position of the detected feature with respect to the own vehicle is represented by the coordinate Zc of the detected feature in the optical axis direction of the stereo camera. Coordinates Zc are d, the parallax, which is the difference in the horizontal position of the detected feature, which is commonly captured in each of the parallax images captured at the same time, b, the base line distance, which is the distance between a pair of cameras, and The focal length of each camera is expressed as fu, and is expressed by equation (1). In this formula, the distance b between the pair of cameras and the focal length fu are known values determined at the time of design, so the distance between the detected feature and the host vehicle is represented by Zc from the parallax d.

Ｓ１６０で状態推定装置１は、地物の対応付けを行う。具体的には、車両が予測位置に存在するものとして、検出地物に含まれる検出標識及び検出境界と地物データに含まれる標識データ及び形状データに含まれる境界データとの位置関係の比較を行う。比較により導出される車両との相対的な位置関係によって、検出標識及び検出境界それぞれが同一物に該当すると推定される標識データ及び境界データの示す標識及び境界を特定することで対応付けを行う。

In S160, the state estimation device 1 associates a feature. Specifically, assuming that the vehicle exists at the predicted position, a comparison of the positional relationship between the detection sign and the detection boundary included in the detected feature and the boundary data included in the marker data and shape data included in the feature data is performed. Do. Based on the relative positional relationship with the vehicle derived from the comparison, the sign and the boundary indicated by the sign data and the boundary data that are assumed to correspond to the same object as the detected sign and the detected boundary are associated with each other.

  In S170, the state estimation device 1 corrects the vertical position. Specifically, the position along the traveling direction of the own vehicle is defined as a vertical position, and the vertical position of the detection marker with respect to the predicted position specified from the map data acquired in S120 and the detection marker measured by the state estimation device 1 in S150. Is calculated as the vertical correction amount. The state estimation device 1 corrects the vertical position represented by the predicted position by the vertical correction amount and sets the corrected vertical position as the comparison position.

In S180, the state estimation device 1 detects the marker center position, which is the center position of the detected marker, by performing the center position detection process. Details of the center position detection processing will be described later.
In S190, the state estimating apparatus 1 determines whether or not the detection is in an inappropriate state as a result of the center position detection processing in S180. If it is determined that the state is inappropriate for detection, the state estimation device determines in S200 whether or not a shape feature that can be used for estimating the road boundary is detected in S140. In S200, when it is determined that a shape feature that can be used for estimating a road boundary is detected, the process ends. That is, the process ends without correcting the lateral position and the vehicle direction. In S200, if it is determined that a shaped feature that can be used for estimating the road boundary has not been detected, the process from S210 is performed. FIG. 6 shows an example of a state in which a shape feature that can be used for estimating a road boundary is not detected as the detected feature. That is, as an example of the state in which the detection is not performed, assuming that the detection range of the optical sensor is inside the two solid lines R1, the detection range is blocked by a running vehicle as indicated by R2 indicated by a dotted line. There are cases. Such a case is exemplified as a case where the road boundary B which is a shaped feature is not detected.

In S210, it is determined whether or not the vertical correction amount calculated in S170 is equal to or greater than a predetermined threshold.
If the state estimation device 1 determines in S210 that the vertical correction amount calculated in S170 is larger than the predetermined threshold, the state estimation device 1 reacquires map data in S220. That is, the state estimation device 1 acquires the map data including the sign data and the boundary data at the comparison position set in S170.

  In S230, the state estimation device 1 re-associates the feature. Specifically, the process is the same as that in S160, except that the vehicle is present at the comparison position set in S170 and the positional relationship is compared.

  If it is determined in S210 that the vertical correction amount corrected by the state estimating device 1 in S170 is smaller than a predetermined threshold, the state estimating device 1 omits the processes of S220 and S230, and performs the processes of S240 and thereafter. I do.

  In S240, the state estimation device 1 corrects the horizontal position. Specifically, this is performed as follows. The lateral position of the detected sign represented by the sign data corresponding to the detected sign with respect to the comparison position or the predicted position is defined as the predicted lateral position, and the lateral position of the detected sign determined based on the image acquired in S130 relative to the own vehicle is defined as the detected lateral position. . This is performed by comparing the predicted lateral position with the detected lateral position. Here, the label data corresponding to the detection marker is the label data corresponding to the detection marker associated in S160 or the detection marker reassociated in S230. The horizontal position based on the image is obtained as follows. That is, the lateral position Xc of the detection marker with respect to the own vehicle is the coordinate Zc of the detection marker in the optical axis direction in the stereo camera, the coordinate u on the image of the detection marker, the horizontal coordinate cu of the image center, and the focal length fu of the stereo camera. From Equation (2).

予測横位置と検出横位置との差分を車両の横補正量とする。状態推定装置１は、当該横補正量だけ比較位置又は予測位置の横位置を車両の横方向にシフトさせることにより、横位置を補正する。

The difference between the predicted lateral position and the detected lateral position is defined as the lateral correction amount of the vehicle. The state estimating device 1 corrects the lateral position by shifting the lateral position of the comparison position or the predicted position in the lateral direction of the vehicle by the lateral correction amount.

  In S250, the state estimation device 1 corrects the vehicle direction. That is, the correction of the vehicle direction is performed as follows. At the comparison position or the predicted position, a virtual image of a boundary display around the own vehicle captured by the optical sensor is estimated based on the boundary data. A comparison is made between the boundary of the road in the virtual image and the boundary of the road in the image captured by the actual optical sensor, and the direction of the vehicle is corrected so that the boundary of each road matches. The method of correcting the vehicle direction is a known technique, and a description thereof will be omitted.

S110 is a state prediction unit, S120 is a map acquisition unit, S140 and S200 are feature detection units, S150 is a relationship detection unit, S160 and S230 are association units, S170 is a first correction unit, S180 is a center detection unit, S210 corresponds to the correction calculation unit, and S240 and S250 correspond to the processing by the second correction unit. Further, among the feature detection units, the one that detects the sign corresponds to the sign detection unit, and the one that detects the boundary corresponds to the boundary detection unit.
<Center position detection processing>
FIG. 7 shows the center position detection process performed by the state estimation device 1.

In S310, the state estimation device 1 acquires the image acquired in S130.
In S320, the state estimation device 1 acquires the detected feature acquired in S140.
In S330, the state estimation device 1 determines whether there is a defect on both the left and right sides of the detection sign. Here, the defect means that a part of the image as shown in FIG. 8 is not correctly recognized by the optical sensor. Here, a state in which a sign is deviated from the angle of view in the detection of a terrestrial object by the optical sensor and complementation is required is described as an example. The loss is not limited to this, but refers to a case where the optical sensor does not properly detect a terrestrial object due to factors such as dirt and an environment to be detected. Specifically, the loss method is performed by the state estimation device 1 comparing the width of the detected sign with the width indicated by the sign data. As a result of the comparison, when the difference between the respective widths is equal to or smaller than the predetermined width threshold, the state estimation device 1 determines that there is no loss on the side of the detection sign.

If the state estimation device 1 determines in S330 that there is no loss, in S340, the midpoint of the width of the detected sign is determined as the sign center position, and the process ends.
If the state estimation device 1 determines in S330 that there is a loss, it determines in S350 whether the loss exists only on one side of the detection sign. Specifically, the state estimation device 1 compares the detected image of the detected sign with the image of the detected sign indicated by the sign data.

  If the state estimating apparatus 1 determines in S350 that there is a defect on only one side of the detected sign by comparing the images, the state estimating apparatus 1 uses the image representing the contour shape of the sign indicated by the sign data to determine in S360 the missing portion. Complement of. That is, since the sign data includes the detection sign of the state without loss, the state estimation device 1 complements the loss by superimposing the sign data on the mark with the loss, and the state estimation device 1 in S370, The midpoint of the width of the complemented detection marker is determined as the marker center position.

  If the state estimation device 1 determines in S350 that there is a defect on both sides of the detection sign by comparing the images, the state estimation device 1 determines in S380 whether the support of the detection sign is detected. . The support | pillar here is a support | pillar which supports a detection sign, and is located in the center of the horizontal direction of a detection sign. That is, the state estimation device 1 compares the support of the detection sign with the support indicated by the sign data. When the strut indicated by the sign data is detected by the comparison, that is, when the strut of the detection sign is detected by the state estimating device 1 in S380, the state estimating device 1 determines the position of the strut as the sign center position in S390. Confirm and end the process.

  If it is not determined in S380 that the support of the detection sign has been detected by the state estimation device 1, it is determined in S400 that the detection is inappropriate, and in S410 the middle point of the width of the detection sign is determined regardless of the presence or absence of the defect. Is determined as the marker center position, and the process ends.

[1-3. effect]
According to the first embodiment described in detail above, the following effects can be obtained.
(1a) According to the state estimation device of the present embodiment, when the vertical correction amount that is the vertical position correction amount is large, the horizontal position and the vehicle direction of the own vehicle are corrected based on the corrected vertical position of the own vehicle. Thus, even when the vertical position is corrected, it is possible to suppress the lateral position and the vehicle direction of the own vehicle from shifting.

  That is, as shown in FIG. 9, it is assumed that the feature M1 is detected at the position P of the own vehicle before the correction. When the feature data D1 corresponding to the feature M1 exists in the map data, the feature data D1 corresponding to the detected feature M1 is changed while maintaining the relative positional relationship between the vehicle and the feature M1. The vertical position is corrected by performing correction so as to overlap. It is assumed that the position of the own vehicle is corrected to the position Q by the correction. Here, when the vertical correction amount is small, the horizontal position of the own vehicle and the vehicle are corrected based on the feature data D2 before correction. By doing so, the time required for reacquisition of data can be reduced. Further, the processing load required for reacquisition of data can be reduced. On the other hand, when the vertical correction amount is large, as shown in FIG. 10, after performing the same vertical position correction as described above, it is assumed that the own vehicle is at the position Q and re-acquisition of the feature data is performed. The horizontal position and the vehicle direction of the own vehicle are corrected by associating with the feature data D3. If the vertical correction amount is large, it is predicted that the deviation of the lateral position and the vehicle azimuth of the own vehicle between the position P before correction and the position Q after correction is large, so that the data is reacquired and associated. Thereby, the prediction accuracy of the state estimation of the own vehicle can be improved.

  (1b) According to the state estimating device of the present embodiment, when the detection is in an inappropriate state, the lateral position and the vehicle direction are corrected using the detected center position detected depending on whether or not the road boundary is detected. Or whether to end the process without performing the correction. Accordingly, in a state where the road boundary cannot be detected, the position of the own vehicle can be corrected based on the sign center position. On the other hand, in a state where the boundary of the road can be detected, by correcting the position of the own vehicle based on the center position of the sign, erroneous correction of the position of the own vehicle can be suppressed. Therefore, the accuracy of correcting the position of the own vehicle can be improved.

  (1c) According to the state estimating device of the present embodiment, by comparing the center positions of the detected signs in the acquired images, it is determined whether or not the image to be corrected is accurately obtained. I do. By changing whether or not to perform correction in accordance with the determination, the accuracy of state estimation is improved.

[2. Other Embodiments]
Although the embodiments of the present disclosure have been described above, the present disclosure is not limited to the above embodiments, and can be implemented with various modifications.

  (2a) In the above embodiment, the optical sensor 5 is a stereo camera. In the above embodiment, the optical sensor 5 is a stereo camera. However, the form of the optical sensor 5 is not limited to this. For example, it may be a laser radar or a monocular camera. When a laser radar is used as the optical sensor 5, the distance between the vehicle and the sign may be measured based on the time from when the laser radar emits laser to when the laser radar receives the laser. In this case, the distance between each feature and the vehicle may be obtained based on a ranging point group, which is a set of points whose distance has been measured, instead of the image captured in the above embodiment.

  (2b) In the above embodiment, the distance between the vehicle and the feature is measured by triangulation using a stereo camera. However, the method for measuring the distance between the vehicle and the feature is not limited to this. is not. For example, the sign data included in the feature data may include feature height data indicating the height of the feature or feature size data indicating the size of the feature. Thereby, the state estimation device 1 compares the feature height data or feature size data with the detected feature height H or feature size S in the image captured by the optical sensor 5 shown in FIG. Thus, the distance between the vehicle and the feature may be measured. In FIG. 11, the sign existing on the side of the road is the detected feature, but the type of the detected feature is not limited to this. It is sufficient that the height H and the size S of the guardrail or the like are stored in advance as feature height data and feature size data. In this case, when the distance is measured based on the height H of the feature, the brightness of the feature portion of the acquired image is saturated, the brightness of the feature portion of the acquired image is high, or the feature of the feature is not present. Even when the size S cannot be measured accurately, the distance between the vehicle and the feature can be measured. Further, when measuring the distance based on the size S of the feature, even when the relative height H between the vehicle and the feature cannot be accurately measured due to a steep road surface or the like, the own vehicle and the feature cannot be measured. Distance can be measured. Further, the distance between the vehicle and the feature may be measured by combining the method for measuring the distance based on the height H of the feature and the method for measuring the distance based on the size S of the feature. Alternatively, a configuration may be adopted in which a method of measuring these distances is selected according to a situation around the vehicle. Specifically, it may be configured to select a distance measuring method based on information acquired from the optical sensor 5 and information on the surrounding environment and road shape included in the map data. The process of selecting the distance measurement method may correspond to the selection unit, and may be performed in S150 of the state estimation process.

  (2c) In the above embodiment, the feature for measuring the distance to the own vehicle is a sign, but the feature for measuring the distance to the own vehicle is not limited to this. For example, a traffic light may be used.

  (2d) A plurality of functions of one component in the above embodiment may be realized by a plurality of components, or one function of one component may be realized by a plurality of components. . Also, a plurality of functions of a plurality of components may be realized by one component, or one function realized by a plurality of components may be realized by one component. Further, a part of the configuration of the above embodiment may be omitted. Further, at least a part of the configuration of the above-described embodiment may be added to or replaced with the configuration of another above-described embodiment. Note that all aspects included in the technical idea specified by the language described in the claims are embodiments of the present disclosure.

  (2e) In addition to the above-described state estimating apparatus 1, a system including the state estimating apparatus 1 as a component, a program for causing a computer to function as the state estimating apparatus 1, and a non-transitional device such as a semiconductor memory storing the program. The present disclosure can be realized in various forms, such as an actual recording medium and a state estimation method.

REFERENCE SIGNS LIST 1 state estimation device 2 own vehicle 3 GNSS receiver 4 map memory 5 optical sensor 6 state quantity sensor 7 control device 11 CPU 12 memory

Claims (11)

A state estimating device (1) mounted on a vehicle and repeatedly estimating a vehicle state including a position and a traveling direction of the vehicle for each observation cycle having a preset cycle,
A state prediction unit (S110) configured to generate a predicted state that is the vehicle state in the current observation cycle, which is predicted using the estimated state that is the vehicle state estimated in the previous observation cycle;
According to a predicted position that is the position of the vehicle included in the predicted state, map data including feature data representing a feature existing around the predicted position and shape data representing a road shape is configured to be acquired. A map acquisition unit (S120);
For each of the observation cycles, based on the detection result of the optical sensor (5), a position feature that is a feature whose installation position is at least registered in the feature data and a shape feature that defines a road shape are extracted. A feature detection unit (S140, S200) configured to detect;
A relation detecting unit (S150) configured to detect a relative positional relation between the position feature and the vehicle;
Assuming that the vehicle exists at the predicted position, the relative position relationship between the position feature and the vehicle detected by the relationship detection unit, the position feature estimated to be the same thing and the An associating unit (S160, S230) configured to associate with the feature data;
A difference between a relative position of the position feature with respect to the vehicle and a relative position of a position represented by the feature data associated with the position feature by the association unit with respect to the predicted position is calculated as a correction amount. A correction calculation unit (S210) configured as
The position along the traveling direction of the vehicle is defined as a vertical position, and the relative position of the position feature with respect to the vehicle and the feature data associated with the position feature by the association unit with respect to the predicted position are represented. A first estimation configured to estimate at least the vertical position of the vehicle in the current measurement cycle by correcting the predicted position using the difference between the position and the relative position as a correction amount and using the correction amount. (S170),
When the vertical correction amount that is the vertical position correction amount among the correction amounts calculated by the correction calculation unit is equal to or less than a preset correction amount threshold value, the predicted position is set as a comparison position, and the vertical correction amount is set. Is larger than the correction amount threshold, when the vertical position of the predicted position is corrected by the vertical correction amount as the comparison position, the shape data at the comparison position acquired by the map acquisition unit By the alignment of the road shape represented by and the road shape indicated by the shape feature detected by the feature detection unit, a lateral position that is a position along the vehicle width direction of the vehicle and a traveling direction of the vehicle A second estimator (S240, S250) configured to estimate at least one of a certain vehicle direction;
A state estimating device comprising: the estimation result in the first estimation unit and the second estimation unit being the estimation state in the current observation cycle. The state estimation device according to claim 1,
The feature detection unit further includes a sign detection unit (S140, S200) configured to detect a sign as the location feature,
The relation detection unit sets a position along the vehicle width direction of the vehicle as a horizontal position and sets a center detection unit (S180) that determines a sign center position that is the center of the sign horizontal position detected by the sign detection unit. The state estimating apparatus further comprising a mark center position detected by the center detection unit as a lateral position of the position feature. The state estimation device according to claim 2,
The feature data includes data of a width of the sign,
The center detection unit, the difference between the width of the sign detected by the sign detection unit and the data of the width of the sign included in the feature data is less than or equal to the width threshold that is a predetermined width threshold. A state estimating device configured to detect a center position of the position feature as the marker center position when it is determined that there is a position feature. The state estimation device according to claim 2,
The feature data includes an image representing the contour shape of the sign,
When the center detection unit determines that the position feature is missing by comparing the sign detected by the sign detection unit and the image, the missing is based on the image. A state estimating device configured to complement a part and determine the marker center position. The state estimation device according to claim 2,
The state estimation device, wherein the center detection unit is configured to set the support to be a marker center position when the support of the marker is recognized by the marker detection unit. The state estimation device according to claim 2, wherein:
The feature detection unit further includes a boundary detection unit configured to detect a boundary display indicating a boundary of a road and a boundary of a lane as the shape feature, and the second estimation unit is configured by the sign detection unit In the case where the sign is not detected, if the boundary of the road can be detected by the boundary detection unit, at least one of the estimation of the lateral position of the vehicle and the estimation of the traveling direction of the vehicle is prohibited. A configured state estimation device. The state estimation device according to any one of claims 1 to 6, wherein:
When the optical sensor is a stereo camera,
The state estimation device, wherein the relation detection unit is configured to measure a distance by triangulation using the stereo camera. The state estimation device according to claim 1, wherein:
A state where the optical sensor is a laser radar, wherein the relation detection unit is configured to measure a distance between the vehicle and the sign by a time from emitting a laser to receiving the laser; Estimation device. The state estimation device according to any one of claims 1 to 6, wherein:
The map data further includes feature size data representing a feature size,
The relationship detection unit is configured to measure a distance between the vehicle and the sign by comparing the feature size data acquired by the map acquisition unit with a size of the sign detected by the sign detection unit. , State estimation device. The state estimation device according to any one of claims 1 to 6, wherein:
The map data further includes feature height data indicating a height at which the feature is installed,
A state estimating device configured to measure a distance between the vehicle and the feature by comparing the feature height data with a height of the feature measured by the optical sensor; . The state estimation device according to any one of claims 1 to 7, wherein:
The relationship detection unit, a plurality of relationship detection unit configured to detect the distance to the feature in a mutually different manner,
A selection unit (S150) configured to select the relationship detection unit to be used in accordance with a surrounding situation of the vehicle.

Images