100. The autonomous driving control device 100 is an example of “mobile object control device”.

[Configuration of Autonomous Driving Control Device 100]

FIG. 4 is a functional configuration diagram of the first control unit 120 and the second control unit 160 included in the autonomous driving control device 100. The first control unit 120 includes, for example, a recognition unit 130 and an action plan generation unit 140. The action plan generation unit 140 includes, for example, a target trajectory generation unit 142, a curvature estimation unit 144, a curvature correction unit 146, and a speed plan generation unit 148.

The recognition unit 130 uses information input from the camera 10 and the radar device 12 via the object recognition device 16 to recognize the state of an object near the subject vehicle M such as a position, a speed, and an acceleration. In particular, the recognition unit 130 acquires the point cloud P representing road division lines from the object recognition device 16, and recognizes a region surrounded by the acquired point cloud P as a lane (travel lane) in which the subject vehicle M is traveling.

The target trajectory generation unit 142 generates a target trajectory TT indicating a path along which the subject vehicle M is to travel in the future based on the point cloud P acquired from the object recognition device 16. FIG. 5 is a diagram for describing an example of the method of generating the target trajectory TT by the target trajectory generation unit 142. First, the target trajectory generation unit 142 sets, as a start point, a point P1 closest to the subject vehicle M of the point cloud P representing a road division line on one side, and sets, as an end point, a point P2 closest to the subject vehicle M of the point cloud P representing a road division line on the other side, to thereby obtain a vector V1 from the start point to the end point. In the example of FIG. 5, the target trajectory generation unit 142 sets, as a start point, a point P1 closest to the subject vehicle M of the point cloud P representing a road division line on the left side, and sets, as an end point, a point P2 closest to the subject vehicle M of the point cloud P representing a road division line on the right side, to thereby obtain a vector V1 from the start point to the end point. However, the vector V1 may be an inverse vector.

Next, the target trajectory generation unit 142 acquires a point P3 second closest to the subject vehicle M among the point cloud P representing the road division line on the left side and the point cloud P representing the road division line on the right side. In FIG. 5, the target trajectory generation unit 142 acquires a point P3 second closest to the subject vehicle M of the point cloud P representing the road division line on the left side. Next, the target trajectory generation unit 142 calculates a unit vector V2′ of a vector V2 from the point P1 serving as a start point to the point P3 serving as an end point, and rotates the calculated unit vector V2′ by −90 degrees to obtain a unit vector V2″. The target trajectory generation unit 142 acquires an orthogonal projection vector V3 of the vector V1 with respect to the unit vector V2″, and estimates the magnitude of the orthogonal projection vector V3 as a width w of the road on which the subject vehicle M is traveling. In other words, the target trajectory generation unit 142 calculates an orthogonal projection vector of the vector V1 with respect to the unit vector in the normal direction of the road division line passing through the point P1.

Next, the target trajectory generation unit 142 determines which of the point cloud P representing the road division line on the left side and the point cloud P representing the road division line on the right side has a higher recognition accuracy. More specifically, for example, the target trajectory generation unit 142 determines which of the point cloud P representing the road division line on the left side and the point cloud P representing the road division line on the right side has a higher recognition accuracy based on whether or not tracking of the point cloud P is successful, the number of times of recognizing the point cloud P, and the length of the line connecting the point cloud P. In the situation illustrated in the left part of FIG. 5, it is detected that the line connecting the point cloud P representing the road division line on the right side is longer than the line representing the road division line on the left side, and thus the target trajectory generation unit 142 determines that the point cloud P representing the road division line on the right side has a higher recognition accuracy.

Next, as illustrated in the right part of FIG. 5, the target trajectory generation unit 142 generates the target trajectory TT by offsetting half the length w/2 of the estimated road width w in the normal direction of the road division line with the road division line on the side of higher recognition accuracy serving as a reference. More specifically, the target trajectory generation unit 142 acquires points TP, which are obtained by offsetting the points of the point cloud P on the side of higher recognition accuracy in the normal direction of the road division line by half the length w/2 of the estimated road width w, and connects the points TP to generate the target trajectory TT.

In the example of FIG. 5, description has been given of a case in which both of the point cloud P representing the road division line on the left side and the point cloud P representing the road division line on the right side are recognized. However, if either one of the point cloud P representing the road division line on the left side and the point cloud P representing the road division line on the right side has failed to be recognized, the point cloud P acquired immediately before failure of recognition and the width w of the road w estimated at that time are used to generate the target trajectory TT by using the above-mentioned method with the point cloud P on the other side serving as a reference. As a result, it is possible to generate the target trajectory TT even if either one of the point cloud P representing the road division line on the left side and the point cloud P representing the road division line on the right side has failed to be recognized.

The curvature estimation unit 144 estimates the curvature of the target trajectory TT generated by the target trajectory generation unit 142. More specifically, the curvature estimation unit 144 estimates the curvature of the target trajectory TT by calculating the curvature of each point TP of the target trajectory TT according to the following expression (1) when the front/back direction of the subject vehicle M is defined as an x axis, the left/right direction of the subject vehicle M is defined as a y axis, and the distance from the subject vehicle M is represented by s. The curvature estimation unit 144 calculates expression (1) by using a numerical analysis technique such as a central difference scheme. In expression (1), κ(s) represents the curvature of the point TP at the position of the distance s from the subject vehicle M, and R(s) represents the curvature radius of the point TP at the position of the distance s from the subject vehicle M. As described above, the point cloud P representing the road division line in the bird's-eye view coordinate system is set at equal intervals, and thus each point TP of the target trajectory TT obtained by offsetting the point cloud P is also set at equal intervals, with the result that a numerical analysis technique such as a central difference scheme can be applied.

κ
⁡
(
s
)

=

1

R
⁡
(
s
)

=

❘
"[LeftBracketingBar]"

dx
ds

⁢

d
2

⁢
y

ds
2

-

dy
ds

⁢

d
2

⁢
x

ds
2

❘
"[RightBracketingBar]"

{

(

dx
ds

)

2

+

(

dy
ds

)

2

}

3
2

(
1
)

FIG. 6 is a graph showing an example of the estimated curvature EC of the target trajectory TT calculated by the curvature estimation unit 144. In FIG. 6, the solid line indicated by EC represents the estimated curvature of the target trajectory TT estimated by the curvature estimation unit 144, and the dotted line indicated by TC represents the true curvature of the road in which the subject vehicle M is traveling. In this manner, the estimated curvature EC of the target trajectory TT set based on the point cloud P converted from the camera coordinate system into the bird's-eye view coordinate system tends to have a larger error as the distance from the subject vehicle M becomes larger. As a result, if the subject vehicle M is decelerated according to the speed plan based on the estimated curvature EC, the occupant of the subject vehicle M my feel strange or uncomfortable for excessive deceleration of the subject vehicle M.

In view of the above-mentioned circumstances, the curvature correction unit 146 corrects the estimated curvature EC by applying a correction filter for correcting the curvature to the estimated curvature EC estimated by the curvature estimation unit 144. More specifically, the curvature correction unit 146 applies a correction filter for setting the upper limit value of the estimated curvature EC so as to decrease the upper limit value of the curvature at the point TP as the distance between the subject vehicle M and the point TP of the target trajectory TT becomes larger.

FIG. 7 is a graph showing an example of the corrected curvature CC of the target trajectory TT corrected by the curvature correction unit 146. In FIG. 7, the two-dotted line indicated by CF represents the correction filter, and the solid line indicated by CC represents the corrected curvature CC of the target trajectory TT corrected by the curvature correction unit 146. As shown in FIG. 7, when the distance s from the subject vehicle M falls within a predetermined value d1, the value of the estimated curvature EC falls below a set value of the correction filter CF, and the value of the corrected curvature CC matches the value of the estimated curvature EC.

On the other hand, when the distance s from the subject vehicle M exceeds the predetermined value d1, the value of the estimated curvature EC exceeds the set value of the correction filter CF, and thus the curvature correction unit 146 applies the correction filter CF to limit the value of the estimated curvature EC to the upper limit value C1 and obtain the corrected curvature CC. In this manner, it is possible to correct only the estimated curvature EC that is expected to have a larger error by setting the upper limit value of the estimated curvature EC smaller as the distance between the subject vehicle M and the point TP of the target trajectory TT becomes larger.

The speed plan generation unit 148 generates a speed plan indicating the future target speed of the subject vehicle M based on the corrected curvature CC corrected by the curvature correction unit 146. More specifically, the speed plan generation unit 148 generates such a speed plan that the lateral acceleration of the subject vehicle M is equal to a predetermined value G that does not cause the occupant to feel uncomfortable under the condition of the corrected curvature CC. The speed plan generation unit 148 deforms the equation of the lateral acceleration v²/R=v²κ=G to generate a speed plan such that the subject vehicle M travels at a speed indicated by the following expression (2). In this embodiment, for the sake of convenience, the horizontal acceleration is calculated by constructing an equation such that the horizontal acceleration of the subject vehicle M is equal to the predetermined value G. However, the predetermined value G in this case is not limited to a specific value, and the predetermined value G may be set to any value within the range that prevents the occupant from feeling uncomfortable or strange.

v
=

const
κ

(
2
)

FIG. 8 is a diagram illustrating an example of the speed plan generated by the speed plan generation unit 148. The speed plan shown in FIG. 8 is generated by the speed plan generation unit 148 for the corrected curvature CC shown in FIG. 7. As shown in FIG. 8, when the distance s from the subject vehicle M follows within the range of the predetermined value d1, the target speed of the subject vehicle M decreases as the distance s from the subject vehicle M becomes larger. This is because as the distance s from the subject vehicle M becomes larger, the corrected curvature CC shown in