FORM 2
THE PATENTS ACT, 1970
(39 of 1970)
&
THE PATENTS RULES, 2003
COMPLETE SPECIFICATION
[See section 10, Rule 13]
STEERING CONTROL DEVICE;
MITSUBISHI ELECTRIC CORPORATION, A CORPORATION ORGANISED AND
EXISTING UNDER THE LAWS OF JAPAN, WHOSE ADDRESS IS 7-3,
MARUNOUCHI 2-CHOME, CHIYODA-KU, TOKYO 1008310, JAPAN
THE FOLLOWING SPECIFICATION PARTICULARLY DESCRIBES THE
INVENTION AND THE MANNER IN WHICH IT IS TO BE PERFORMED

2
DESCRIPTION
FIELD OF THE INVENTION
[0001]
The present application relates to the field of a steering control device.
5 BACKGROUND OF THE INVENTION
[0002]
A steering control device which is equipped with a steering support device for
reducing the steering force of a driver is proposed. Further, a steering control device
which is equipped with, in addition to the steering support device, an automatic steering
10 device that automatically steers based on the shape of a road is proposed.
For example, in a vehicle steering device that uses both of a power steering device
and an automatic steering device which are disclosed in the Patent Document 1, when a
driver carries out an override during the automatic steering operation, automatic steering
control will deviate from a target value. Thereby, an electric motor is actuated so that
15 the difference between a target value and an actual value may be reduced. In
consequence, there is shown a problem that steering by a driver is hindered.
[0003]
In order to solve this problem, in the Patent Document 1, a technology is proposed
which is provided with a means to detect a driver torque by a driver and the shape of a
20 road on which a vehicle travels, an electric motor which generates an additional driver
torque, a means to determine a driver support torque according to the driver torque, a
means to determine an automatic driver torque according to the shape of a road, a means
to determine an amount of target state according to the shape of a road, a means to
detect an amount of actual state, a means to estimate an amount of vehicle state by the

3
automatic driver torque, a means to compute an amount of comparison state which is
computed as a weighted average between the amount of actual state and the amount of
estimated state, using a weight which is determined based on the driver torque, and a
means to control the output of an electric motor based on the sum of the driver torque
5 and the automatic driver torque, wherein the automatic driver torque is determined
according to the deviation between the amount of target state and the amount of
comparison state.
[0004]
According to the contents which are proposed here, simultaneous operations in the
10 steering support mode and the automatic steering mode can be achieved.
[0005]
That is, in the Patent Document 1, there is disclosed a technology in which
continuous switching is designed between the control of an electric power steering
device and the control of an automatic steering device, according to the increase or
15 decrease of the information on steering forces. As a result, when the steering force of a
driver is applied, the control ratio of an electric power steering device will increase, and
when the steering force of a driver decreases, the ratio of the automatic steering device
will increase. Thereby, the shift between the electric power steering mode and the
automatic steering mode will be achieved smoothly.
20 CITATION LIST
PATENT LITERATURE
[0006]
Patent Document 1: Japanese Unexamined Patent Application Publication No. 10 -
076967

4
SUMMARY OF THE INVENTION
TECHNICAL PROBLEM
[0007]
In such a steering control device, while the lane keep assist system ( LKAS ) is in
5 operation, if a driver carries out an override, for example, there occurs a subject that the
steering control becomes unstable by the interference between the driver torque by a
driver and the automatic driver torque. That is, because the gain of the automatic driver
torque is directly computed based on the driver torque, the gain also changes suddenly,
if the driver torque changes suddenly, and the automatic driver torque will also change
10 suddenly. The sum of a driver torque, an automatic driver torque, and a driver support
torque becomes an input driver torque, which is applied to the steering for the change of
a rudder. Thereby, if the automatic driver torque changes suddenly, the driver torque
will also change suddenly, for the purpose of compensating the sudden change.
Therefore, if either the driver torque by a driver or the automatic driver torque changes
15 suddenly, there occurs a problem that they interfere mutually and the steering control
becomes unstable.
[0008]
The present application is made in order to solve the above-mentioned problem, and
aims at offering a stable steering control, even in a case where the driver torque by a
20 driver or the automatic driver torque changes suddenly.
SOLUTION TO PROBLEM
[0009]
A steering control device according to the present application, includes;
a driver support torque computing part which computes a driver support torque

5
according to a driver torque,
an automatic driver torque computing part which computes an automatic driver
torque according to road condition, and
an additional driver torque computing part which computes an additional driver
5 torque according to the driver support torque and the automatic driver torque,
wherein the additional driver torque computing part receives an output from at least
one of a first torque correction computing part and a second torque correction
computing part;
wherein the first torque correction computing part includes:
10 a weight computing part which computes an increment of weight based on a
magnitude of the driver torque, and accumulates the increment of weight to generate a
weight,
a clip value computing part which computes an automatic driver torque clip value
according to the weight, and
15 a clip processing part which clip processes the automatic driver torque with the
automatic driver torque clip value, to limit an upper limit value and a lower limit value
thereof, and outputs a clip processed automatic driver torque, to the additional driver
torque computing part, and
the second torque correction computing part includes:
20 a gain computing part which computes an increment of gain according to the driver
torque, and accumulates the increment of gain to generate a gain, and
a gain processing part which outputs an automatic driver torque multiplied by the
gain, to the additional driver torque computing part.
ADVANTAGEOUS EFFECTS OF INVENTION

6
[0010]
According to the present application, at least either the computation of an increment
of weight based on the magnitude of the driver torque, or the computation of an
increment of gain based on the driver torque, will be conducted. Thereby, the weight
5 itself or the gain itself does not change suddenly, even though the driver torque changes
suddenly, and then, it becomes possible to prevent the steering control from becoming
unstable.
BRIEF EXPLANATION OF DRAWINGS
[0011]
10 Fig. 1 is a block diagram which shows the schematic constitution of Embodiment 1.
Fig. 2 is a system constitutional diagram which shows the Embodiment 1.
Fig. 3 is a constitutional diagram which shows an example of the hardware of the
Embodiment 1.
Fig. 4 is an explanatory diagram of the coordinate system of the Embodiment 1.
15 Fig. 5 is a flow chart which shows the Embodiment 1.
Fig. 6 is an explanatory diagram of a map for computing an increment of weight
from a driver torque, in the Embodiment 1.
Fig. 7 is a schematic view of the scene of the steering control.
Fig. 8 is a schematic view which shows the mechanism of the steering control.
20 Fig. 9 is an explanatory diagram which shows a map for performing a gain
computation from the driver torque.
Fig. 10 is a schematic view which shows the mechanism by which the steering
control becomes stable in the Embodiment 1.
Fig. 11 is a schematic view which shows the mechanism by which the hunting of the

7
weight is reduced in Embodiment 2.
Fig. 12 is a schematic view of a scene in which the driver torque converges
regardless of the curvature of a driving lane in Embodiment 3.
Fig. 13 is a schematic view which shows that the reaction force which is applied to a
5 driver is changed, depending on the curvature in the Embodiment 2.
Fig. 14 is a schematic view which shows that the reaction force which is applied to a
driver becomes constant, regardless of the curvature in the Embodiment 3.
Fig. 15 is a block diagram which shows the constitution of Embodiment 4.
Fig. 16 is a flow chart which shows the operation of the Embodiment 4.
10 Fig. 17 is an explanatory diagram of a map for computing a first threshold, from the
degree of a first deviation in the Embodiment 4.
Fig. 18 is a schematic view of a scene in the Embodiment 4, where a larger reaction
force which is applied to a driver can be obtained, as the degree of the first deviation
becomes larger.
15 Fig. 19 is a schematic view which shows that, even though the degree of the first
deviation increases, the reaction force which is applied to a driver does not increase, in
the Embodiment 4.
Fig. 20 is a schematic view of the Embodiment 4 which shows that a larger reaction
force which is applied to a driver can be obtained, as the degree of the first deviation
20 becomes larger.
Fig. 21 is a block diagram which shows the constitution of Embodiment 5.
Fig. 22 is a flow chart which shows the operation of the Embodiment 5.
Fig. 23 is an explanatory diagram of a map for computing an increment of gain from
the driver torque in the Embodiment 5.

8
Fig. 24 is a schematic view which shows that, when the magnitude relationship
between a steering wheel angle and a target steering wheel angle is reversed, the
automatic driver torque changes suddenly.
Fig. 25 is a schematic view which shows that, when the magnitude relationship
5 between a steering wheel angle and a target steering wheel angle is reversed, the
automatic driver torque does not change suddenly.
Fig. 26 is a block diagram which shows the constitution of the Embodiment 6.
Fig. 27 is a flow chart which shows the operation of the Embodiment 6.
Fig. 28 is an explanatory diagram of a map for computing a first threshold from the
10 degree of the first deviation, in the Embodiment 6.
Fig. 29 is a block diagram which shows the constitution of Embodiment 7.
Fig. 30 is a flow chart which shows the operation of the Embodiment 7.
Fig. 31 is a schematic view which shows that, when a LKAS torque clip value is
large, the convergence of the driver torque is slow.
15 Fig. 32 is a schematic view which shows that, when a LKAS torque clip value is
large, the convergence of the driver torque is fast.
Fig. 33 is a flow chart which shows the processing of a weight computing part in
Embodiment 8.
Fig. 34 is an explanatory diagram of a map for computing an increment adjustment
20 coefficient from the degree of a second deviation in the Embodiment 8.
Fig. 35 is an explanatory diagram which shows that the driver torque does not
converge to a predetermined first threshold in the Embodiment 3.
Fig. 36 is an explanatory diagram which shows that the driver torque converges to a
predetermined first threshold in the Embodiment 9.

9
Fig. 37 is an explanatory diagram which shows that the driver torque does not
converge to a predetermined second threshold in the Embodiment 7.
Fig. 38 is an explanatory diagram which shows that the driver torque converges to a
predetermined second threshold in the Embodiment 10.
5 Fig. 39 is a block diagram which shows the constitution of the Embodiment 11.
Fig. 40 is a flow chart which shows the operation of the Embodiment 11.
Fig. 41 is an explanatory diagram which shows that, when the curvature of a driving
lane is constant in the Embodiment 11, the reaction force which is applied to a driver
can be made constant, regardless of the curvature.
10 DESCRIPTION OF EMBODIMENTS
[0012]
Embodiment 1.
Fig. 1 is a block diagram which shows the schematic constitution of a steering
control system of the Embodiment 1. The present steering control system is mounted on
15 a vehicle 1, and is equipped with a steering control unit 200, and a driver torque
acquisition part 110 which is connected to the steering control unit, a vehicle
information acquisition part 120, a lane information acquisition part 130, and a steering
use actuator 310. The steering control unit 200 controls the steering of a vehicle 1
( henceforth referred to as “host vehicle” ) which is provided with the steering control
20 system.
The steering control unit 200 is, for example, an electric power steering electronic
control unit ( EPS-ECU ), and the like.
[0013]
The driver torque acquisition part 110 acquires a driver torque by a driver of the host

10
vehicle 1 ( henceforth “driver torque by a driver” is referred to as “driver torque” ). The
driver torque acquisition part 110 is, for example, a driver torque sensor.
The vehicle information acquisition part 120 acquires vehicle information, including
the speed of a host vehicle, acceleration, direction, angular velocity, and others. The
5 vehicle information acquisition part 120 is, for example, a steering wheel angle sensor, a
yaw rate sensor, a speed sensor, an acceleration sensor, and the like.
The lane information acquisition part 130 acquires lane information, which is the
information on a lane on which the host vehicle 1 travels. It is assumed that the
positions of left and right lane markings on a host vehicle driving lane, or information
10 on the position of a lane center is included in the lane information. The lane information
acquisition part 130 is, for example, a front camera. It is worth noticing that, the lane
information acquisition part 130 can obtain lane information from the Global
Navigation Satellite System ( GNSS ), and map data, or LiDAR and LiDAR use map
data.
15 [0014]
The steering control unit 200 is equipped with a steering control device 201 and a
target steering wheel angle computing part 220.
Further, the steering control device 201 is equipped with a driver support torque
computing part 210, an automatic driver torque computing part 230, a torque correction
20 computing part 201A, and an additional driver torque computing part 270, and the
torque correction computing part 201A is equipped with a weight computing part 240, a
clip value computing part 250, and a clip processing part 260.
[0015]
The driver support torque computing part 210 computes a driver support torque for

11
supporting the steering by a driver, based on the information containing a driver torque
from the driver torque acquisition part 110.
The target steering wheel angle computing part 220 computes a target steering
wheel angle for maintaining the center of a host vehicle driving lane, based on the
5 information from the lane information acquisition part 130. It is worth noticing that,
instead of the lane center, it is allowed to produce a route with due consideration for
obstacles and others, and to compute a target steering wheel angle for following the
route.
[0016]
10 The automatic driver torque computing part 230 computes an automatic driver
torque for making a real steering wheel angle follow the target steering wheel angle
which is computed in the target steering wheel angle computing part 220. It is worth
noticing that, it is allowed to compute directly an automatic driver torque for
maintaining the center of a host vehicle driving lane, without mounting the target
15 steering wheel angle computing part 220.
[0017]
The weight computing part 240 computes a weight based on the information which
contains at least a driver torque. It is worth noticing that, here, the weight indicates a
ratio of the maximum value of the automatic driver torque clip value ( the automatic
20 driver torque clip value which is determined for safety ) and the minimum value ( the
override torque clip value ). In the following explanation, the ratio of allocation will be
explained as a weight.
The clip value computing part 250 computes an automatic driver torque clip value
using the weight.

12
The clip processing part 260 clips an automatic driver torque with an automatic
driver torque clip value, so that the upper limit and lower limit of the automatic driver
torque may be limited by the automatic driver torque clip value.
The additional driver torque computing part 270 computes an additional driver
5 torque based on a driver support torque and a clip processed automatic driver torque.
And, the steering control device controls so that the steering use actuator 310 may
generate the additional driver torque.
The steering use actuator 310 is an actuator to generate an additional driver torque
which is computed in the additional driver torque computing part 270, and, for example,
10 is an EPS motor ( Electric Power Steering motor ).
[0018]
Fig. 2 is a constitutional diagram of a steering control system. The host vehicle 1, on
which the steering control system is mounted, is equipped with a steering wheel 2, a
steering axis 3, a steering unit 4, an EPS motor 5, a driver torque sensor 111, a steering
15 wheel angle sensor 121, a yaw rate sensor 122, a speed sensor 123, an acceleration
sensor 124, a front camera 131, a GNSS sensor 132, a navigation gear 133, a LiDAR
134, a LiDAR use map 135, and the steering control unit 200 which is shown in Fig. 1.
[0019]
The steering wheel 2 is, so called, a handle, which is for a driver to operate the host
20 vehicle 1. The steering wheel 2 is connected to the steering axis 3, and the steering unit
4 is in conjunction with the steering axis 3. The steering unit 4 supports front wheels as
the steering wheel, with rotational movement freedom, and in addition, is supported by
a body frame with steering freedom. Therefore, the driver torque, which is generated by
the driver’s operation of the steering wheel 2, rotates the steering axis 3, and the

13
steering unit 4 changes the steering of the front wheels to a horizontal direction,
according to the rotation of the steering axis 3. In consequence, the driver can operate
the amount of horizontal movement of the host vehicle 1, at the time when the host
vehicle 1 moves forward or backward.
5 It is worth noticing that, the steering axis 3 can be also rotated by the EPS motor 5.
Controlling the current which flows into the EPS motor 5, the steering control unit 200
can change freely the steering of the front wheels, independently of the driver’s
operation of the steering wheel 2. The example of the front wheel steering is shown here.
However, a driver and the steering control unit 200 may conduct the rear wheel steering
10 or the steering of front wheels and rear wheels.
[0020]
The steering control unit 200 is connected with the EPS motor 5, the driver torque
sensor 111, the steering wheel angle sensor 121, the yaw rate sensor 122, the speed
sensor 123, the acceleration sensor 124, the front camera 131, the GNSS sensor 132, the
15 navigation gear 133, the LiDAR 134, and the LiDAR use map 135.
The driver torque sensor 111 detects the driver torque of the steering axis 3 by a
driver. Here, it is assumed that the driver torque acquisition part 110 which is shown in
Fig. 1 is constituted by the driver torque sensor 111.
[0021]
20 The steering wheel angle sensor 121 detects the angle of the steering wheel 2. The
yaw rate sensor 122 detects the yaw rate of the host vehicle 1. The speed sensor 123
detects the speed of the host vehicle 1. The acceleration sensor 124 detects the
acceleration of the host vehicle 1. Here, it is assumed that the vehicle information
acquisition part 120 is constituted by the steering wheel angle sensor 121, the yaw rate

14
sensor 122, the speed sensor 123, and the acceleration sensor 124.
[0022]
The front camera 131 is installed at a position where a lane marking ahead of the
vehicle can be detected as a picture, and detects the front environment of the host
5 vehicle, such as lane information and the position of an obstacle, based on picture
information. Here is shown only a front camera which detects the front environment.
However, a camera which detects the environment of the back or the side is also
allowed to be installed in the host vehicle 1. Here, it is assumed that the lane
information acquisition part 130 which is shown in Fig. 1 is constituted by the front
10 camera 131.
[0023]
The GNSS sensor 132 receives electric waves from positioning satellites by an
antenna, and conducts the computation for positioning. Thereby, the GNSS sensor 132
outputs the absolute position and absolute direction of the host vehicle 1. The navigation
15 gear 133 has the function to compute an optimal driving route to a place to go ( a
destination ) which is set by a driver, and keeps in memory the map data including road
information on respective roads which constitute a road network. The road information
is map node data which represent road alignments. In addition, each of the map node
data includes the absolute position ( latitude, longitude, and altitude ) of each node, lane
20 width, Kant angle, tilt angle information, and the like. Here, it is assumed that the lane
information acquisition part 130 which is shown in Fig. 1 is constituted by the GNSS
sensor 132 and the navigation gear 133.
[0024]
The LiDAR 134 irradiates a laser, and detects the reflective wave, and thereby,

15
detects a circumferential object, on the basis of the host vehicle 1. The LiDAR use map
135 is a map which is created based on the detection results of the LiDAR, and the
estimation of a host position can be attained, by combining the map with the detection
results of the LiDAR 134. Moreover, in the LiDAR use map 135, road information
5 which is likely to be included in the navigation gear 133 is stored in memory, and the
position of the host vehicle 1 within a driving lane can be acquired by combining the
road information with the result of the host position estimation. Here, it is assumed that
the lane information acquisition part 130 which is shown in Fig. 1 is constituted by the
LiDAR 134 and the LiDAR use map 135. The steering control unit 200 is an integrated
10 circuit, such as a microprocessor. As an example of the hardware is shown in Fig. 3, the
steering control unit is equipped with a processor 50, such as an A-D conversion circuit,
a D-A conversion circuit, and a central processing unit ( CPU ), and a memory storage
51, such as a Read Only Memory ( ROM ) and a Random Access Memory ( RAM ).
The processor 50 performs processing on the information which is input from each of
15 the sensors, according to the program which is stored in the memory storage 51. The
processor 50 controls so that the EPS motor 5 may generate the additional driver torque.
Although the contents of the memory storage 51 are not illustrated, the memory storage
is equipped with volatile memory storages, such as random access memories, and
non-volatile auxiliary memory storages, such as flash memories. Moreover, the memory
20 storage can be equipped with auxiliary memory storages of hard disk type, instead of
flash memories. The processor 50 achieves the execution of the program which is input
from the memory storage 51. In this case, the program is input into the processor 50
from the auxiliary memory storages through the volatile memory storages. Moreover,
the processor 50 may output the data of operation results and the like, to the volatile

16
storages of the memory storage 51, and may save the data through the volatile storages
in the auxiliary memory storage.
[0026]
Fig. 4 is a drawing schematically showing a coordinate system which is used in the
5 Embodiment 1. Symbols x and y of Fig. 4 are of a host vehicle coordinate system where
the center of gravity of the host vehicle is set on the origin point, and the front of the
host vehicle is taken as x-axis, and the left-hand direction is taken as y-axis. In this Fig.
4, the symbol e θ indicates an angle ( a directional error ) between the lane center at a
host vehicle position and the x-axis. The symbol e0 indicates a distance from the lane
10 center to the host vehicle ( a lateral position at a host vehicle position ), where the lane
center is shown by the curve A. The symbol e Ld indicates a distance ( a lateral position
at a look-ahead distance ) from the lane center to the look-ahead distance ( Point B in
the drawing ). It is worth noticing that, when the steering control unit 200 is equipped
with a route generation equipment, it is allowed to use a target route instead of the lane
15 center.
Fig. 5 is a flow chart which shows the procedure of the steering control device of the
Embodiment 1.
In Step S110 of Fig. 5, a driver torque T Driver by a driver is acquired in the driver
torque acquisition part 110.
20 [0027]
In Step S120 of Fig. 5, vehicle information is acquired in the vehicle information
acquisition part 120. The vehicle information is the information of steering wheel angle
of a host vehicle, yaw rate, speed, acceleration, and the like. In the present Embodiment,
steering wheel angle δ, yaw rate γ, speed V, and acceleration α are acquired.

17
[0028]
In Step S130 of Fig. 5, lane information is acquired in the lane information
acquisition part 130. The lane information is the information on the right and left lane
markings of a lane, which a host vehicle should travel, and the information on a lane
5 center. In the present Embodiment, coefficients at the time when the right and left lane
markings are represented by a third-order polynomial equation in a host vehicle
coordinate system will be acquired.
[0029]
That is, with regard to the left lane marking, values of C l0 to C l3 of the
10 following Equation 1 will be acquired.
[ Equation 1 ]
Eq. １
( 1 )
With regard to the right lane marking, values of C r0 to C r3 of the following
15 Equation 2 will be acquired.
[ Equation 2 ]
Eq. ２
( 2 )
At this time, with regard to the lane center, values of C C0 to C C3 of the following
20 Equation 3 will be acquired.
[ Equation 3 ]
Eq. ３
( 3 )
Here, the coefficients C l3, C r3, and C C3 indicate the estimated curvature change

18
rates of a route, and the coefficients C l2, C r2, and C C2 indicate the curvature
components of a route, and the coefficients C l1, C r1, and C C1 indicate the yaw angle
components of a route, with regard to a host vehicle, and the coefficients C l0, C r0, and
C C0 indicate the position components in the horizontal direction of a route, with regard
5 to a host vehicle. And each of the coefficients satisfies the relation which is shown in
the Equation 4.
[ Equation 4 ]
Eq. ４
( 4 )
10 [0030]
At this time, the lateral position e 0 at a host vehicle position, the lateral position e Ld
at a look-ahead distance, and the directional error e θ, which are shown in Fig. 3, are
defined by the following equations respectively.
[ Equation 5 ]
15 Eq. 5
( 5 )
[ Equation 6 ]
Eq. ６
( 6 )
20 [ Equation 7 ]
Eq. ７
( 7 )
However, a look-ahead distance is represented as L d. It is worth noticing that,
information on lane markings is not limited to the third-order polynomial equation, and

19
may be represented by any function.
[0031]
Next, in Step S210 of Fig. 5, a driver support torque T Assist for assisting the steering
of a driver is computed in the driver support torque computing part 210. To compute the
driver support torque T Assist 5 , publicly known techniques will be used, where, for
example, computations are performed based on the driver torque T Driver and the speed V
of a host vehicle.
[0032]
Next, in Step S220 of Fig. 5, a target steering wheel angle δ Ref for making a host
10 vehicle follow a lane center is computed in the target steering wheel angle computing
part 220. Publicly known techniques, such as PID, linear secondary regulator, and
model prediction control, will be used for the computation of the target steering wheel
angle δ Ref.
It is worth noticing that, when the steering control unit 200 is equipped with a route
15 generation part, it is allowed to compute a target steering wheel angle for following a
target route. In this Embodiment 1, using the lateral position e Ld at a look-ahead
distance ( Point B ) and the directional error e θ, like the Equation 8, the target steering
wheel angle δ Ref will be computed.
[ Equation 8 ]
20 Eq. 8
+ + ( 8 )
Here, the symbol t is a variable indicating time.
[0033]
Next, in Step S230 of Fig. 5, an automatic driver torque T Auto for making the

20
steering wheel angle δ follow the target steering wheel angle δ Ref is computed in the
automatic driver torque computing part 230.
For the operation of the automatic driver torque, publicly known techniques, such as
PID, linear secondary regulator, and model prediction control, will be used. In this
5 Embodiment 1, using the steering wheel angle δ and the target steering wheel angle δ Ref,
like the Equation 9, the automatic driver torque T Auto will be computed.
[ Equation 9 ]
Eq. 9
+ ( 9 )
10 [0034]
When a driver is driving a vehicle and is releasing his hands from the steering wheel,
it is desirable that the steering wheel angle follows the target steering wheel angle with
sufficient accuracy. Therefore, a proportionality gain K 4 is set usually as a large value
( for example, 10 Nm/deg or so ).
15 Next, in Step S240 of Fig. 5, a weight w is computed in the weight computing part
240.
In the case where the smaller the weight w is, the smaller the automatic driver
torque clip value T Clip, Auto becomes, an increment of weight Δw will be computed, so
that, if the magnitude of the driver torque | T Driver | is smaller than a predetermined first
20 threshold θ W ( for example, 0.5 Nm or so ), the increment of weight Δw may become
positive, and, if the magnitude of the driver torque | T Driver | is larger than a first
threshold θ W, the increment of weight Δw may become negative. Regarding the
computation of the increment of weight Δw, it is allowed to use a map which is in
accordance with the magnitude of the driver torque | T Driver |, or to use a constant value.

21
[0035]
For example, when computing Δw with a map M W ( | T Driver | ）of | T Driver |, the
weight w will be computed like the following Equations.
[ Equation 10 ]
5 Eq. 10
( 10 )
[ Equation 11 ]
Eq. 11
( 11 )
10 [ Equation 12 ]
Eq. 12
( 12 )
Here, the symbol k is a variable which represents a discretized time, and the symbol
t samp is a computation cycle of the weight w. It is worth noticing that, in the case where
15 the larger the weight w is, the automatic driver torque clip value T Clip, Auto becomes
smaller, the increment of weight Δw will be computed so that, if the magnitude of driver
torque | T Driver | is smaller than the first threshold θ W, the increment of weight Δw may
become negative, and if the magnitude of driver torque | T Driver | is larger than the first
threshold θ W, the increment of weight Δw may become positive. In the present
20 Embodiment, it is assumed that, the smaller the weight w is, the smaller the automatic
driver torque clip value T Clip, Auto becomes. Moreover, the range of the weight w is
restricted so that w ∈ [0, 1] may be satisfied in the Equation 12, however, this range
may not be [0, 1].
[0036]

22
In this way, the weight w is computed through the computation of the increment of
weight Δw. Thereby, comparing the case where the weight w is computed directly from
the magnitude of the driver torque | T Driver |, the weight w is less likely to change
suddenly, even when the magnitude of the driver torque | T Driver | changes suddenly. In
5 addition, it is possible to prevent the steering control from becoming unstable.
Moreover, by setting a small increment of weight Δw near the first threshold θ W, the
weight w and the automatic driver torque clip value T Clip, Auto become less likely to
carry out hunting, at the time when the magnitude of the driver torque | T Driver | is near
the first threshold θ W. In addition, the benefit obtained here is that the steering control
10 is less likely to become unstable. It is worth noticing that, the first threshold θ W is not a
fixed value, and may be a variable value.
[0037]
Next, in Step S250 of Fig. 5, the automatic driver torque clip value T Clip, Auto is
computed in the clip value computing part 250. For example, the automatic driver
15 torque clip value T Clip, Auto is computed as a weighted average of two kinds of clip
values, using the weight w, as follows.
[ Equation 13 ]
Eq. 13
( 13 )
20 Here, the symbol T Clip and LKAS ( > T Clip, OVR ) is a LKAS torque clip value, which is
set for safety in LKAS. Usually, in order to enable the driving even on a sharp curve, the
LKAS torque clip value is set to be about 10 times larger value ( for example, 20 Nm or
so ), compared with the driver torque, which a driver can generate. The symbol T Clip,
OVR ( ≥ 0 ) is an override torque clip value, and is set as a value ( for example, 1 Nm or

23
so ) whose magnitude is almost as large as a driver torque, which a driver can generate.
[0038]
By computing the automatic driver torque clip value T Clip, Auto in this way, the
automatic driver torque clip value T Clip, Auto becomes equal to the LKAS torque clip
5 value T Clip, LKAS, when the weight w is a maximum value. In addition, when the weight
w is a minimum value, the automatic driver torque clip value T Clip, Auto becomes equal
to the override torque clip value T Clip, OVR.
That is, it can be said that, the LKAS torque clip value T Clip, LKAS is a parameter
which defines the maximum value of the automatic driver torque clip value T Clip, Auto,
10 and in addition, the override torque clip value T Clip, OVR is a parameter which defines the
minimum value of the automatic driver torque clip value T Clip, Auto.
[0039]
And, when a driver is releasing his hands from the steering wheel, an autonomous
driving system can generate an automatic driver torque which is required for the lane
15 keeping. In addition, the benefit obtained here is that the autonomous driving system
can restrict the automatic driver torque, to such an extent that the steering of a driver
may not be disturbed during an override. It is worth noticing that, concerning the
method of computing the automatic driver torque clip value T Clip, Auto, it becomes
unnecessary to compute the automatic driver torque clip value T Clip, Auto as a weighted
20 average, if the automatic driver torque clip value T Clip, Auto is computed so that, when
the weight w is a maximum value, the automatic driver torque clip value T Clip, Auto may
become equal to the LKAS torque clip value T Clip, LKAS; and when the weight w is a
minimum value, the automatic driver torque clip value T Clip, Auto may become equal to
the override torque clip values T Clip, OVR; and when the weight w is a value other than

24
those values, the automatic driver torque clip value T Clip, Auto may decrease
monotonically.
[0040]
Next, in Step S260 of Fig. 5, a clip processing of the automatic driver torque T Auto
5 is performed in the clip processing part 260. The clip processing is performed as follows,
using the automatic driver torque clip value T Clip, Auto.
[ Equation 14 ]
Eq. 14
( 14 )
10 [0041]
Next, in Step S270 of Fig. 5, an additional driver torque T EPS is computed in the
additional driver torque computing part 270. For example, the additional driver torque T
EPS is computed as the sum of the automatic driver torque T Auto and the support driver
torque T Assist, like the following.
15 [ Equation 15 ]
Eq. 15
( 15 )
Or, using the weight w, the additional driver torque T EPS may be computed as a
weighted average, as follows.
20 [ Equation 16 ]
Eq. 16
( 16 )
Or, in addition to these methods, as long as based on the automatic driver torque T
Auto and the support driver torque T Assist, any method can be applied to compute the

25
additional driver torque T EPS.
[0042]
Next, in Step S310 of Fig. 5, the steering control device controls so that the steering
use actuator 310 may generate the additional driver torque T EPS. Publicly known
5 techniques are used for the control of the steering use actuator.
Fig. 6 is a drawing which shows an example of a map M W, for computing the
increment of weight Δw in the weight computing part 240. The map is designed so that,
when the magnitude of the driver torque | T Driver | is smaller than the first threshold θ W,
the increment of weight Δw may become positive, and when | T Driver | is larger than the
10 first threshold θ W, Δw may become negative.
[0043]
The reduction amount is made larger than the increase amount. Thus, the benefit
obtained here is that when a driver carries out an override, the automatic driver torque
clip value can be decreased quickly. Moreover, the absolute value of the increment of
15 weight Δw is made small near the first threshold θ W. Thus, the benefit obtained here is
that when | T Driver | is near the first threshold θ W, the weight w and the automatic driver
torque clip value T Clip, Auto become less likely to carry out hunting.
[0044]
Fig. 7 is a schematic view which shows an example of the scene in which the
20 steering control is likely to become unstable. A host vehicle X is travelling on a straight
road, and LKAS is in operation. For that reason, the autonomous driving system is
performing the steering control so that the vehicle may travel on a lane center. And,
vehicles Y parked on a street are lined up on the left front all the way for about 100 m,
and it is assumed that an autonomous driving system cannot recognize vehicles Y

26
parked on a street. In that case, a driver carries out an override, and tries to drive with an
offset of about 1 m on the right side of a driving lane. In Fig. 7, the dashed line C shows
the target route of autonomous driving, and the solid line D shows the target route of the
driver.
5 [0045]
Fig. 8 is a schematic view which shows a phenomenon in which, when the gain K of
the automatic driver torque is directly adjusted according to the driver torque T Driver, the
steering control becomes unstable in the scene of Fig. 7. The relationship among the
lateral position, the torque, the steering wheel angle, and the gain is represented in this
10 Fig. 8. It is worth noticing that, in the drawing showing the relationship of torques, the
dashed line T1 shows an automatic driver torque, the solid line T2 shows an input
torque, and the solid line T3 shows a driver torque by a driver. Moreover, in the drawing
showing the steering wheel angle, the solid line S1 shows a steering wheel angle, and
the dashed line S2 shows a target steering wheel angle. Moreover, in the drawing
15 showing the gain, the solid line G represents the change state of the gain. Here, a case
can be considered in which, when the driver torque T Driver increases, the gain K
decreases as shown in Fig. 9. It is worth noticing that, in the following explanation, it is
simply assumed that the support driver torque T Assist is 0 Nm at any time. Therefore, the
additional driver torque T EPS becomes equal to the product of the gain K and the
automatic driver torque T Auto. In addition, the input torque T Input 20 which is applied to a
steering axis for the change of a rudder becomes the sum of the driver torque T Driver and
the product of the gain K and the automatic driver torque T Auto. That is, the following
equation 17 is satisfied.
[ Equation 17 ]

27
Eq. 17
( 17 )
[0046]
Next, explanation will be made about the mechanism by which the steering control
5 becomes unstable in Fig. 8. First, in the vicinity of 5 s ( at Time E ) of the time axis of
Fig. 8, a driver carries out an override, and, in order to move a vehicle to the right side
of a driving lane, the driver torque of a right direction is generated. Then, because the
steering wheel rotates to the right, the autonomous driving system generates an
automatic driver torque of the left direction, in order to return to the left. However,
10 because the gain decreases by the increase of the driver torque, the automatic driver
torque also decreases. After that, as the vehicle approaches straight going, a driver
reduces the driver torque, in order to return the steering wheel to 0 deg. However, since
the driver torque decreases, the gain increases.
[0047]
15 At this time, the gain is raised by the amount equal to the decreased driver torque.
Thereby, if the reduction amount of the driver torque is large, the increase amount of the
gain will also become large. Due to the rapid increase of the gain, influence of the
automatic driver torque, which returns a vehicle to a lane center, increases sharply. Then,
the driver, who wishes to maintain an offset drive at Time F, increases the driver torque
20 rapidly, in order to cancel the automatic driver torque. Then, since the automatic driver
torque decreases rapidly again, the driver torque which is required to cancel it also
increases sharply. Henceforth, the automatic driver torque and the driver torque will
repeat a rapid increase and a rapid decrease. Thereby, as shown by G in Fig. 8, the
driver torque, the gain, and the automatic driver torque perform vibrational behaviors,

28
and the steering control becomes unstable. In this way, one of the causes which induce
unstable steering control is the direct computation of the gain which is conducted based
on the driver torque.
[0048]
5 It is difficult to solve this subject by parameter tuning. That is, when a driver is
releasing his hands from the steering wheel, the autonomous driving system needs to
generate an automatic driver torque which is required for the lane keeping, and, during
an override, the autonomous driving system tries to decrease the gain of the automatic
driver torque, to such an extent that the gain may not interfere with the steering of a
10 driver. In this case, it is necessary to set the gain as 1 at 0 Nm, like the map of Fig. 9,
and to decrease the gain to near zero, at a driver torque whose magnitude is easy for a
driver to generate ( for example, 2.5 Nm ). Usually, in order to enable the lane keeping
even at a sharp curve, the maximum value of the automatic driver torque ( for example,
50 Nm ) is about 10 times as large as the maximum value of the driver torque which a
15 driver can generate ( for example, 5 Nm ). For that reason, the fluctuation range of the
automatic driver torque due to the fluctuation of the gain is much larger than the
fluctuation range of the driver torque. No matter how the shape of the map of Fig. 9 is
changed, a vibration phenomenon among the driver torque and the gain and the
automatic driver torque, like the one described above, cannot be avoided.
20 [0049]
As a comparative example, Fig. 9 is a drawing which shows an example of the map
which will be used when the gain of the automatic driver torque is directly adjusted
according to the driver torque. In Fig. 9, when the driver torque is 0.5 Nm or less, the
gain is 1. After that, the gain decreases monotonously until the driver torque is up to 2.5

29
Nm, and when the driver torque is 2.5 Nm or more, the gain becomes 0.
[0050]
Fig. 10 is a schematic view which shows that the steering control can be stabilized
in the scene of Fig. 7. For confirmation, it is assumed that the first threshold θ W = 0.5
5 Nm, the LKAS torque clip value T Clip, LKAS = 20 Nm, and the override torque clip value
T Clip, OVR = 0.0 Nm. That is, if the magnitude of the driver torque exceeds 0.5 Nm, the
weight decreases from 1 to 0, by the Equations 10, 11, and 12, and the automatic driver
torque clip value also decreases from 50 Nm to 0.0 Nm. Moreover, if the magnitude of
the driver torque becomes less than 0.5 Nm, the weight increases from 0 to 1, and the
10 automatic driver torque clip value also increases from 1.0 Nm to 50 Nm. Moreover, also
in Fig. 10, it is assumed that the driver support torque is 0 Nm at any time.
[0051]
It is worth noticing that, in the drawing showing the relationship of torques, the
dashed line T1 shows an automatic driver torque, the solid line T2 shows an input
15 torque, the solid line T3 shows a driver torque by a driver, the fine dashed line T4 shows
an automatic driver torque clip value, and the symbol Th1 shows a first threshold θ W.
Moreover, in the drawing showing steering wheel angles, the solid line S1 shows a
steering wheel angle, and the dashed line S2 shows a target steering wheel angle.
Moreover, in the drawing showing the weight, the dashed line W represents the change
20 state of the weight.
[0052]
Next, explanation will be made about the mechanism by which the steering control
is unlikely to become be unstable in Fig. 10. First, in Fig. 10, like in Fig. 8, in the
vicinity of 5 s ( at Time E ) of the time axis, the driver torque of right direction is

30
generated, in order that a driver may move a vehicle to the right side of a driving lane.
Then, when the driver torque exceeds the first threshold of 0.5 Nm, the weight
decreases from 1 to 0. At the same time, the automatic driver torque clip value decreases
from 50 Nm to 0.0 Nm. As a result, the automatic driver torque is also clipped at 1.0
5 Nm. After that, as the vehicle approaches straight going, a driver reduces the driver
torque, in order to return the steering wheel to 0 deg.
[0053]
At this time, even though the driver torque becomes less than the first threshold of
0.5 Nm, as long as the increment of weight is set small as shown in Fig. 6, the weight
10 will not increase rapidly. Thereby, the automatic driver torque clip value does not
increase rapidly, nor does the clipped automatic driver torque increase rapidly. After that,
when going straight ahead, in order to maintain the steering wheel at 0 deg, a driver
generates a driver torque so that the automatic driver torque may be cancelled. However,
as shown in Fig. 10, the automatic driver torque converges to a fixed value. And, when
15 an override torque clip value is larger than a first threshold θ W which is determined
beforehand, the convergence value becomes a value equal to an override torque clip
value. In addition, when the override torque clip value is smaller than the predetermined
first threshold θ W, the convergence value becomes a value equal to the predetermined
first threshold θ W.
20 [0054]
In the below, the reason is explained. Since the steering control device needs to
maintain the steering wheel at 0 deg during the offset driving of a straight road, the sum
of the driver torque and the automatic driver torque is 0 Nm, when the support driver
torque is 0 Nm. For that reason, in order to have a converged driver torque, the

31
automatic driver torque also needs to be converged. Supposing that the automatic driver
torque is clipped at the automatic driver torque clip value, both of the automatic driver
torque clip value and the weight need to be converged. There are three kinds of patterns
in which the weight converges. The first kind is a pattern in which the increment of
5 weight is positive, and the weight converges to 1, the second kind is a pattern in which
the increment of weight is negative and the weight converges to 0, and the third kind is
a pattern in which the increment of weight is 0, and the weight converges to a suitable
value. Since the weight must be on the decrease during an override, patterns of the
second kind or the third kind can be considered.
10 [0055]
Since the increment of weight needs to be 0 in order to have a convergence by the
pattern of third kind, the driver torque needs to become equal to a predetermined first
threshold θ W. Therefore, the convergence value of the driver torque becomes a value
which is equal to the predetermined first threshold θ W. However, when the override
15 torque clip value is larger than the predetermined first threshold θ W, the automatic
driver torque is converged to an override torque clip value which is larger than the
predetermined first threshold θ W. Since cancellation of this automatic driver torque
requires a driver torque which is larger than the predetermined first threshold, the
increment of weight becomes negative and the convergence occurs by the pattern of
20 second kind.
[0056]
That is, the weight is converged to 0. From the reason mentioned above, when an
override torque clip value is larger than a predetermined first threshold θ W, the
convergence value of the driver torque during an offset driving becomes a value which

32
is equal to an override torque clip value. In addition, when the override torque clip value
is smaller than the predetermined first threshold θ W, the convergence value becomes a
value which is equal to the predetermined first threshold θ W. In this way, according to
the constitution of the present Embodiment, it becomes possible to predict the
5 convergence value of the driver torque during an offset driving. Thereby, the benefit
obtained here is that it is easy to adjust the reaction force which a driver receives from
the steering wheel.
[0057]
In this way, according to the constitution of the Embodiment 1, since the weight is
10 computed through the increment of weight, the weight itself does not change suddenly,
even though the driver torque changes suddenly. Therefore, since the automatic driver
torque clip value also does not change suddenly, the automatic driver torque also does
not change suddenly. Then the benefit obtained here is that it is possible to prevent the
steering control from becoming unstable. Moreover, the convergence value of the driver
15 torque during an offset driving can be predicted according to the magnitude relationship
of the override torque clip value and the first threshold θ W which is determined
beforehand. Therefore, the benefit obtained here is that it is easy to adjust the reaction
force which a driver receives from the steering wheel.
[0058]
20 It is worth noticing that, although the LKAS is assumed in the Embodiment 1, it is
allowed to employ Lane Departure Prevention System ( LDPS ), instead of the LKAS.
Moreover, the increment of weight is computed based on the driver torque, and a
clip value is changed in the Embodiment 1. However, if the increment of gain is
computed instead of the increment of weight and the gain of the automatic driver torque

33
is changed, the sudden change of the automatic driver torque can be prevented at the
time when the driver torque changes suddenly. Then, the benefit obtained here is that it
is possible to prevent the steering control from becoming unstable. Or, it is allowed to
use both the change of a clip value by the weight and the change of the gain. As a result,
5 the wider range of tuning becomes available.
[0059]
Embodiment 2.
In the Embodiment 1, set is only the condition that the override torque clip value
( the minimum value of the automatic driver torque clip value ) is smaller than the
10 LKAS torque clip value ( the maximum value of the automatic driver torque clip value ).
However, it is allowed to add the condition that the override torque clip value is larger
than the maximum value of the predetermined first threshold θ W. As a result, the driver
torque during an offset driving is converged to a value which is larger than the
predetermined first threshold θ W, the weight is converged to a minimum value, and the
15 automatic driver torque is converged to the minimum value of the automatic driver
torque clip value. Therefore, the convergence of the driver torque, the weight, and the
automatic driver torque is accelerated, and the stability of the steering control can be
improved.
[0060]
20 In the below, explanation will be made about the Embodiment 2. Explanation which
overlaps with the Embodiment 1 is omitted here.
In the Embodiment 2, the override torque clip value is set to be larger than the
maximum value of the first threshold θ W, which is determined beforehand. When the
override torque clip value is smaller than the first threshold θ W, the driver torque is

34
converged to the first threshold θ W, as explained in the Embodiment 1. At this time, the
increment of weight may carry out hunting between the positive and the negative, in the
vicinity of the first threshold θ W, the weight also changes and the convergence becomes
slower. On the other hand, by setting the override torque clip value to be larger than the
5 maximum value of the first threshold θ W, the driver torque is converged, during an
offset driving, to a value which is larger than the first threshold θ W. Therefore, the
increment of weight becomes always negative, and the weight does not change, and the
convergence becomes faster.
[0061]
10 Fig. 11 is a schematic view which shows that, in the scene which is shown in Fig. 7,
the convergence of the driver torque becomes faster in the Embodiment 2. It is assumed
that, the first threshold θ W = 0.5 Nm, the LKAS torque clip value T Clip, LKAS = 20 Nm,
and the override torque clip value T Clip, OVR = 1.0 Nm.
It is worth noticing that, in the drawing which shows the relationship of torques, the
15 dashed line T1 shows an automatic driver torque, the solid line T2 shows an input
torque, the solid line T3 shows a driver torque by a driver, the fine dashed line T4 shows
an automatic driver torque clip value, and the symbol Th1 shows a first threshold θ W.
Moreover, in the drawing showing steering wheel angles, the solid line S1 shows a
steering wheel angle, and the dashed line S2 shows a target steering wheel angle.
20 Moreover, in the drawing showing the weight, the dashed line W represents the change
state of the weight.
[0062]
In Fig. 11, the driver torque is converged to the minimum value of the automatic
driver torque clip value, i.e., an override torque clip value. Since the override torque clip

35
value is over the predetermined first threshold θ W, the increment of weight becomes
always negative and the weight converges to 0. Compared with Fig. 10 in which the
increment of weight carries out hunting between the positive and the negative and the
weight fluctuates, the convergence of the weight becomes faster.
5 In this way, it is not a mere change of the design to set an override torque clip value
which is larger than the maximum value of the first threshold θ W. It is a necessary
condition for the convergence value of the weight to be converged to 0 at the time of an
offset driving.
[0063]
10 It is worth noticing that, in Fig. 10, it is assumed that the driver support torque is 0
Nm at any time. However, the driver support torque is not 0 Nm, in fact. For that reason,
the driver torque which is required for the cancellation of the automatic driver torque
becomes lower by the amount equal to the driver support torque. Therefore, in order to
make a weight and an automatic driver torque clip value converge to a minimum value
15 during an offset driving, the influence of the driver support torque also needs to be taken
into consideration, and it is required that the minimum value of the automatic driver
torque clip value is made larger than the predetermined first threshold, with a margin
between the two.
[0064]
20 In this way, according to the constitution of the Embodiment 2, the driver torque
during an offset driving is converged to a value which is larger than the predetermined
first threshold, the weight is converged to a minimum value, and the automatic driver
torque is converged to the minimum value of the automatic driver torque clip value.
Thereby, the convergence of the driver torque, the weight, and the automatic driver

36
torque is accelerated, and the benefit obtained here is that the stability of the steering
control can be improved.
[0065]
Embodiment 3.
5 In the Embodiment 2, a condition is added that the override torque clip value ( the
minimum value of the automatic driver torque clip value ) is larger than the maximum
value of the first threshold, which is determined beforehand. Instead, it is allowed to add
a condition that the override torque clip value is smaller than the minimum value of the
first threshold θ W. As a result, since the driver torque during an offset driving is
10 converged to the first threshold θ W, regardless of the curvature of a driving lane, the
reaction force which is applied to a driver can be made constant, regardless of the
curvature.
[0066]
In the below, explanation will be made about the Embodiment 3. Explanation which
15 overlaps with the Embodiments 1 and 2 is omitted here.
Fig. 12 is a schematic view which shows an example of a scene in which, in the
present Embodiment, the driver torque is likely to converge to a predetermined first
threshold θ W, regardless of the curvature of a driving lane. The host vehicle X is
travelling on a straight road which is connected to a left curve, and the LKAS is in
20 operation. For that reason, the autonomous driving system is performing steering
control so that the vehicle may travel on a lane center. And, vehicles Y parked on a
street are lined up all the way from a straight road to the left curve at a left front, and it
is assumed that the autonomous driving system cannot recognize vehicles parked on a
street. At this time, a driver carries out an override, and tries to drive the host vehicle X

37
with an offset of about 1 m on the right side of a driving lane. It is worth noticing that,
in Fig. 12, the dashed line C shows the target route of the autonomous driving, and the
solid line D shows the target route of a driver.
[0067]
5 Fig. 13 is a schematic view which shows that, in the scene of Fig. 12, when the
override torque clip value is not set to be smaller than the minimum value of the first
threshold θ W, which is determined beforehand, the reaction force which is applied to a
driver depends on the curvature. For confirmation, it is assumed that the predetermined
first threshold θ W = 0.5 Nm, the LKAS torque clip value T Clip, LKAS = 20 Nm, and the
10 override torque clip value T Clip, OVR = 1.0 Nm. In Fig. 13, for the same reason as in Fig.
11, the driver torque is converged, in the straight section, to the minimum value of the
automatic driver torque clip value, i.e., the override torque clip value. When a vehicle
enters a relief section toward a left curve, the input torque T2 needs to be raised by the
amount equal to a self aligning torque. However, since the automatic driver torque T1
15 cannot be increased due to a clip, the driver torque T3 is instead raised in a left
direction.
[0068]
It is worth noticing that, in the drawing showing the relationship of torques, the
dashed line T1 shows an automatic driver torque, the solid line T2 shows an input
20 torque, the solid line T3 shows a driver torque by a driver, the fine dashed line T4 shows
an automatic driver torque clip value, and the symbol Th1 shows a first threshold θ W.
Moreover, in the drawing showing steering wheel angles, the solid line S1 shows a
steering wheel angle, and the dashed line S2 shows a target steering wheel angle.
Moreover, in the drawing showing the weight, the dashed line W represents the change

38
state of the weight.
[0069]
When the driver torque reaches the first threshold θ W, the weight will increase.
Thereby, both of the automatic driver torque clip value and the automatic driver torque
5 also will increase. As a result, the input torque is raised by the amount equal to the
increased self aligning torque. After that, the driver torque is converged to the first
threshold θ W, also in the left curve section. In this way, when the override torque clip
value T Clip, OVR is set to be larger than the maximum value of the predetermined first
threshold θ W, the convergence value of the driver torque will change depending on the
10 curvature, and the reaction force which is applied to a driver will also change depending
on the curvature.
[0070]
Fig. 14 is a schematic view which shows that, in the scene of Fig. 12, a case where
the override torque clip value is set to be smaller than the minimum value of the
15 predetermined first threshold θ W, in other words, in the present Embodiment, the
reaction force which is applied to a driver does not depend on the curvature. For
confirmation, it is assumed that the predetermined first threshold θ W = 0.5 Nm, the
LKAS torque clip value T Clip, LKAS = 20 Nm, and the override torque clip value T Clip,
OVR = 0.0 Nm.
20 [0071]
It is worth noticing that, in the drawing which shows the relationship of torques, the
dashed line T1 shows an automatic driver torque, the solid line T2 shows an input
torque, the solid line T3 shows a driver torque by a driver, the fine dashed line T4 shows
an automatic driver torque clip value, and the symbol Th1 shows a first threshold θ W.

39
Moreover, in the drawing showing steering wheel angles, the solid line S1 shows a
steering wheel angle, and the dashed line S2 shows a target steering wheel angle.
Moreover, in the drawing showing the weight, the dashed line W represents the change
state of the weight.
5 [0072]
In Fig. 14, the driver torque is converged to a first threshold θ W, in a straight section,
for the same reason as in Fig. 10. After that, when the vehicle enters a relief section or a
left curve, the input torque needs to be raised by the amount equal to the self aligning
torque. However, when the driver torque becomes even slightly smaller than the first
10 threshold θ W, the weight will increase. In addition, since the automatic driver torque
clip value and the automatic driver torque also increase, the input torque is on the
increase, while the driver torque is converged to the first threshold θ W. In this way,
when the override torque clip value is set to be smaller than the minimum value of the
predetermined first threshold θ W, the driver torque converges to the first threshold θ W,
15 regardless of the curvature. Therefore, the benefit obtained here is that the reaction force
which is applied to a driver also can be made constant, regardless of the curvature.
[0073]
In this way, it is not a mere change of the design to set an override torque clip value
which is smaller than the minimum value of the first threshold θ W. In addition, it is a
20 necessary condition for the driver torque to converge to the first threshold θ W at the
time of an offset driving. It is worth noticing that, in Fig. 14, the driver support torque is
set to 0 Nm at any time. However, the driver support torque is not 0 Nm, in fact. For
that reason, the driver torque which is required for the cancellation of the automatic
driver torque is reduced by the amount equal to the driver support torque.

40
[0074]
Therefore, in order to have a driver torque which is converged to the first threshold
θ W during an offset driving, the influence of the driver support torque also needs to be
taken into consideration, and it is required that the minimum value of the automatic
5 driver torque clip value is made larger than the predetermined first threshold θ W, with a
margin between the two. In this way, according to the constitution of the present
Embodiment, the driver torque converges to the predetermined first threshold θ W
during an offset driving, regardless of the curvature of a driving lane. Therefore, the
benefit obtained here is that the reaction force which is applied to a driver can be made
10 constant, regardless of the curvature.
[0075]
Embodiment 4.
In the Embodiment 3, added is the condition that the override torque clip value T Clip,
OVR is smaller than the minimum value of the predetermined first threshold θ W.
15 However, it is allowed further to employ a larger first threshold θ W, according to the
magnitude of the degree of a first deviation. The driver torque during an offset driving
converges to a first threshold θ W. Thereby, as the degree of the first deviation becomes
larger, a larger convergence value of the driver torque can be obtained. That is, the
reaction force which is applied to a driver can be increased, and the risk of deviating
20 from a lane during an override can be reduced.
[0076]
In the below, explanation will be made about the Embodiment 4. Explanation which
overlaps with the Embodiments 1, 2, and 3 is omitted here.
Fig. 15 is a block diagram which shows the schematic constitution of the steering

41
control system of the Embodiment 4. Explanation about portions which are common to
Fig. 1 will be omitted.
[0077]
The difference from Fig. 1 is that the torque correction computing part 201A is
5 equipped with a first threshold computing part 241. The first threshold computing part
241 computes a first threshold θ W based on the degree of a first deviation. Here, the
degree of the first deviation is a lateral position at a host vehicle position, for example.
Or, the degree of the first deviation is a lateral position at a look-ahead distance, a
distance to the lane marking of a host vehicle driving lane at a host vehicle position, a
10 distance to the lane marking of a host vehicle driving lane at a look-ahead distance, and
a time until a host vehicle crosses the lane marking of a host vehicle driving lane. In
addition, any variable which indicates the degree of the deviation of a host vehicle may
be used.
[0078]
15 The weight computing part 240 computes a weight based on the first threshold
which is computed in the first threshold computing part 241.
[0079]
Fig. 16 is a flow chart which shows the procedure of the steering control device of
the Embodiment 4. Explanation about portions which are common to Fig. 5 will be
20 omitted.
In Step S241 of Fig. 16, a first threshold θ W is computed in the first threshold
computing part 241. Computation is conducted so that the first threshold θ W may
become larger, as the degree of the deviation becomes larger. For the computation of the
first threshold θ W, a map may be used, or a suitable monotonically increasing function

42
may be used.
[0080]
Next, in Step S240 of Fig. 16, a weight is computed in the weight computing part
240. Except that, the first threshold θ W which is computed in the first threshold
5 computing part 241 is used for the computation of the weight, the weight is computed in
the same procedure as the Step S240 of Fig. 5.
[0081]
Fig. 17 is a drawing which shows an example of the map M θW, for the computation
of the first threshold θ W, based on the degree of the first deviation in the first threshold
10 computing part 241. Here, it is assumed that the absolute value of a lateral position | e0 |
at a host vehicle position is the degree of a first deviation. This map is designed so that
the first threshold θ W may monotonically increase, according to the absolute value of a
lateral position | e0 |. Since the driver torque during an offset driving converges to the
first threshold θ W, a larger convergence value of the driver torque can be obtained, as
15 the degree of the first deviation becomes larger. That is, the reaction force which is
applied to a driver can be increased, and the benefit obtained here is that the risk of
deviating from a lane during an override is reduced.
[0082]
It is worth noticing that, as explained in the Embodiments 1 and 3, in order to make
20 the driver torque during an offset driving converge to the first threshold θ W, the
override torque clip value T Clip, OVR needs to be smaller than the minimum value of the
predetermined first threshold θ W. Therefore, the map M θW is designed so that the first
threshold θ W may always become larger than the override torque clip value T Clip, OVR.
Moreover, in the present Embodiment, the absolute value of a lateral position | e0 | is

43
used as the input of the map. However, it is allowed to use | e0 | which is multiplied by a
coefficient, where the coefficient becomes 1 at the outer side of a curve, and -1 at the
inner side of a curve. And, when the Map M θW includes a horizontal axis which is
designed to a negative domain, it becomes possible to change convergence values of the
5 driver torque at the outer side and inner side of a curve.
[0083]
Fig. 18 is a schematic view of the present Embodiment 4 which shows an example
of a scene where, as the lateral position at a host vehicle position is larger, the reaction
force which is applied to a driver is likely to become larger. The host vehicle X is
10 travelling on s straight road, and LKAS is in operation. For that reason, the autonomous
driving system is performing the steering control so that the host vehicle may travel on a
lane center. And, the vehicle group Y1 which is parked on a street is lined up on the left
front all the way for about 100m, and it is assumed that the autonomous driving system
cannot recognize the vehicle group Y1 which is parked on a street. At this time, a driver
15 carries out an override, and drives the host vehicle, at first, with an offset of about 0.5 m
on the right side of the driving lane. From the middle of the way, corresponding to the
vehicle group Y2 which is parked on a street, the driver tries to drive with an offset of
about 1 m. It is worth noticing that, in Fig. 18, the dashed line C shows the target route
of the autonomous driving, and the solid line D shows the target route of the driver.
20 [0084]
Fig. 19 is a schematic view which shows that, when the first threshold θ W is a fixed
value in the scene of Fig. 18, the reaction force which is applied to a driver does not
change, even though the lateral position at a host vehicle position is increased. For
confirmation, it is assumed that the predetermined first threshold θ W = 0.5 Nm, the

44
LKAS torque clip value T Clip, LKAS = 20 Nm, and the override torque clip value T Clip,
OVR = 0.0 Nm.
It is worth noticing that, in the drawing showing the relationship of torques, the
dashed line T1 shows an automatic driver torque, the solid line T2 shows an input
5 torque, the solid line T3 shows a driver torque by a driver, the fine dashed line T4 shows
an automatic driver torque clip value, and the symbol Th1 shows a first threshold θ W.
Moreover, in the drawing showing steering wheel angles, the solid line S1 shows a
steering wheel angle, and the dashed line S2 shows a target steering wheel angle.
Moreover, in the drawing showing the weight, the dashed line W represents the change
10 state of the weight.
[0085]
A driver carries out an override at around 7 s of the time axis of Fig. 19, and the
lateral position is increased to 0.5 m or so. At this time, since the driver torque is
converged to 0.5 Nm of the first threshold θ W, the reaction force which is applied to a
15 driver becomes 0.5 Nm. After that, the driver carries out steering again at around 17 s,
and the lateral position is increased to 1 m or so. However, since the first threshold θ W
is fixed at 0.5 Nm, the driver torque converges to 0.5 Nm, and the reaction force which
is applied to a driver remains at 0.5 Nm. In this way, when the first threshold θ W is a
fixed value, the reaction force which is applied to a driver does not increase, even
20 though a lateral position is increased. Then, the driver cannot perceive, through the
reaction force, that the degree of the deviation is on the increase.
[0086]
Fig. 20 is a schematic view which shows that, when a larger first threshold θ W is
employed according to the magnitude of the degree of the first deviation in the scene of

45
Fig. 18, the reaction force which is applied to a driver is increased, if the lateral position
at a host vehicle position is increased. For confirmation, it is assumed that the
predetermined first threshold θ W is computed using the map of Fig. 17. Moreover, it is
assumed that the LKAS torque clip value T Clip, LKAS = 50 Nm, and the override torque
5 clip value T Clip, OVR = 0.0 Nm.
It is worth noticing that, in the drawing showing the relationship of torques, the
dashed line T1 shows an automatic driver torque, the solid line T2 shows an input
torque, the solid line T3 shows a driver torque by driver, the fine dashed line T4 shows
an automatic driver torque clip value, and the symbol Th1 shows a first threshold θ W.
10 Moreover, in the drawing showing steering wheel angles, the solid line S1 shows a
steering wheel angle, and the dashed line S2 shows a target steering wheel angle.
Moreover, in the drawing showing the weight, the dashed line W represents the change
state of the weight.
[0087]
15 In Fig. 20, it can be seen that the first threshold θ W is increasing, with the increase
of the lateral position. As a result, while the driver torque is converged to 1 Nm or so,
when a lateral position is 0.5 m, the driver torque is converged to 2 Nm, when the lateral
position is 1 m. That is, when the lateral position increases, the reaction force which is
applied to a driver will also increase. As a result, the driver can perceive, through the
20 reaction force, that the degree of the deviation is on the increase. For that reason, the
risk of deviating from a lane during an override can be reduced.
[0088]
In this way, according to the constitution of the Embodiment 4, the first threshold θ
W changes depending on the degree of the first deviation. Since the driver torque during

46
an offset driving is converged to the first threshold θ W, a larger convergence value of
the driver torque can be obtained, as the degree of the first deviation becomes larger.
That is, the reaction force which is applied to a driver can be increased, and the risk of
deviating from a lane during an override can be reduced.
5 [0089]
Embodiment 5.
In the Embodiment 1, the clip value of the automatic driver torque is changed
according to the driver torque. However, it is allowed to change a gain, instead of the
clip value. At this time, like the computation of the weight, a gain is computed through
10 the increment of gain. Thereby, even though the driver torque changes suddenly, the
gain itself does not change suddenly, and the stability of the steering control can be
improved. Moreover, when computing the automatic driver torque based on the
deviation of a steering wheel angle and a target steering wheel angle, sudden change of
the automatic driver torque can be prevented, when the positive or negative of the
15 deviation of a steering wheel angle is reversed, and the stability of the steering control
improves.
In the below, explanation will be made about the Embodiment 5. Explanation which
overlaps in the Embodiments 1 - 4 is omitted here.
[0090]
20 In the Embodiment 5, a gain is computed, according to the magnitude of the driver
torque, based on the automatic driver torque, and a final automatic driver torque is
computed by multiplying the automatic driver torque by the gain. Here, for the
computation of a gain, a second threshold is determined beforehand. When the driver
torque is small with respect to this second threshold, computation will be performed so

47
that the increment of gain may become positive. In addition, when the driver torque is
larger than the predetermined second threshold, computation will be performed so that
the increment of gain may become negative.
[0091]
5 As a result, since the gain becomes small during an override, the automatic driver
torque can be decreased to such an extent that the steering of a driver may not be
interfered during an override. Further, the benefit obtained here is that the autonomous
driving system can generate the automatic driver torque which is required for the lane
keeping, when the driver is releasing his hands from the steering wheel. Moreover, like
10 the computation of the weight, by computing a gain through the increment of gain, the
gain itself does not change suddenly, even though the driver torque changes suddenly.
Therefore, the automatic driver torque does not change suddenly, and the benefit
obtained here is that it is possible to prevent the steering control from becoming
unstable. Moreover, while the processing for clipping the automatic driver torque is
15 non-linear transformation, the processing for multiplying a gain is linear transformation.
Thereby, the time differentiation value of the automatic driver torque is also subject to
the influence of a gain. As a result, when the gain is small, the time differentiation value
of the automatic driver torque also becomes small. In addition, the benefit obtained here
is that, when the magnitude relationship between the steering wheel angle and the target
20 steering wheel angle is reversed, sudden change of the automatic driver torque can be
prevented.
[0092]
Fig. 21 is a block diagram which shows the schematic constitution of the steering
control system of the Embodiment 5. Explanation about portions which are common to

48
Fig. 1 will be omitted.
The difference from Fig. 1 is the constitution of the torque correction computing
part 201B, and the torque correction computing part 201B is equipped with a gain
computing part 280 and a gain processing part 290.
5 The gain computing part 280 computes a gain based on the information which
contains at least a driver torque.
The gain processing part 290 multiplies the automatic driver torque by the gain.
The additional driver torque computing part 270 computes an additional driver
torque based on a driver support torque and the gain processed driver torque. And, the
10 steering control device controls so that the steering use actuator 310 may generate the
additional driver torque.
[0093]
Fig. 22 is a flow chart which shows the procedure of the steering control device of
the Embodiment 5. Explanation about portions which are common to Fig. 5 will be
15 omitted.
In Step S280 of Fig. 22, a gain K is computed in the gain computing part 280. The
method of computing the gain K is the same as that for the weight. When the magnitude
of the driver torque | T Driver | is smaller than a second threshold θ K ( for example, 0.5
Nm or so ), an increment ΔK is computed so that the increment of gain ΔK may become
positive. In addition, when the magnitude of the driver torque | T Driver 20 | is larger than the
second threshold θ K, an increment ΔK is computed so that the increment of gain ΔK
may become negative. Regarding the computation of the increment ΔK, it is allowed to
use a map which corresponds to the magnitude of the driver torque | T Driver |, or to use a
constant value.

49
[0094]
For example, when computing the increment ΔK with the map M K ( | T Driver｜ ) of
the magnitude | T Driver | of the driver torque, the gain K is computed as follows.
[ Equation 18 ]
5 Eq. 18
( 18 )
[ Equation 19 ]
Eq. 19
( 19 )
10 [ Equation 20 ]
Eq. 20
( 20 )
In the Equation 20, the range of the gain K is restricted so that K∈ [0, 1] may be
satisfied, however, this range may not be [0, 1].
15 [0095]
In this way, the gain K is computed through the computation of the increment of
gain ΔK. Thereby, even when the magnitude of the driver torque | T Driver | changes
suddenly, the gain K becomes less likely to change suddenly, comparing the case, where
the gain K is computed directly from the magnitude of the driver torque | T Driver |. The
20 benefit obtained here is that it is possible to prevent the steering control from becoming
unstable. Moreover, a small increment of gain ΔK is set near the second threshold θ K.
Thereby, when the magnitude of the driver torque | T Driver | is near the second threshold
θ K, a possibility that a gain will carry out hunting can be reduced. In addition, the
benefit obtained here is that steering control is less likely to become unstable. It is worth

50
noticing that, the second threshold θ K is not a fixed value, and may be a variable value.
In Step S290 of Fig. 22, the automatic driver torque is multiplied by a gain, in the
gain computing part 280.
[0096]
5 Fig. 23 is a drawing which shows an example of the map M K for computing the
increment of gain in the gain computing part 280. Here, the map M K is designed so that,
when the magnitude of the driver torque | T Driver | is smaller than the second threshold θ
K, the increment of gain ΔK may become positive, and in addition, when the magnitude
of the driver torque | T Driver | is larger than the second threshold θ K, the increment of
10 gain ΔK may become negative.
The benefit obtained here is that, by having a decrease amount which is larger than
an increase amount, a decreased gain will be obtained quickly, when a driver carries out
an override. Moreover, at the time when the magnitude of the driver torque | T Driver | is
near the second threshold θ K, the increment of gain ΔK is made small. Thereby, a
15 benefit obtained here is that the gain K is less likely to carry out hunting, at the time
when the magnitude of the driver torque | T Driver | is near the second threshold θ K.
[0097]
Fig. 24 is a schematic view which shows that, when the magnitude relationship
between a steering wheel angle and a real steering wheel angle is reversed in the scene
20 of Fig. 7, the automatic driver torque changes suddenly. For confirmation, it is assumed
that the predetermined first threshold θ W = 0.5 Nm, the LKAS torque clip value T Clip,
LKAS = 20 Nm, and the override torque clip value T Clip, OVR = 1.0 Nm.
It is worth noticing that, in the drawing showing the relationship of torques, the
dashed line T1 shows an automatic driver torque, the solid line T2 shows an input

51
torque, the solid line T3 shows a driver torque by a driver, the fine dashed line T4 shows
an automatic driver torque clip value, and the symbol Th1 shows a first threshold θ W.
Moreover, in the drawing showing steering wheel angles, the solid line S1 shows a
steering wheel angle, and the dashed line S2 shows a target steering wheel angle.
5 Moreover, in the drawing showing the weight, the dashed line W represents the change
state of the weight.
In Fig. 24, the magnitude relationship between a steering wheel angle and a target
steering wheel angle is reversed in the vicinity of 8 s ( at Time H ) of the time axis. The
automatic driver torque is computed by the Equation 9. Since the proportionality gain is
10 set as a large value, the automatic driver torque changes suddenly from the maximum
value to the minimum value of the automatic driver torque clip value, immediately after
the magnitude relationship between the steering wheel angle and the target steering
wheel angle is reversed. In this way, when the automatic driver torque is clipped, the
sudden change of the automatic driver torque cannot be prevented within the range of a
15 clip.
[0098]
Fig. 25 is a schematic view which shows that, even though the magnitude
relationship between the steering wheel angle and the real steering wheel angle is
reversed, the automatic driver torque does not change suddenly in the Embodiment 5.
20 For confirmation, it is assumed that the predetermined second threshold θ K = 0.5 Nm.
Moreover, it is assumed that the LKAS torque clip value T Clip, LKAS = 20 Nm is set for
safety.
It is worth noticing that, in the drawing showing the relationship of torques, the
dashed line T1 shows an automatic driver torque, the solid line T2 shows an input

52
torque, the solid line T3 shows a driver torque by a driver, and the symbol Th2 shows a
second threshold θ K. Moreover, in the drawing showing steering wheel angles, the solid
line S1 shows a steering wheel angle, and the dashed line S2 shows a target steering
wheel angle. Moreover, in the drawing showing the gain, the solid line G represents the
5 change state of the gain.
In Fig. 25, the magnitude relationship between the steering wheel angle and the
target steering wheel angle is reversed in the vicinity of 8 s ( at Time H ) of the time axis.
However, since the gain is on the decrease, the time differentiation value of the
automatic driver torque also decreases, and the automatic driver torque does not change
10 suddenly. In this way, by decreasing the gain of the automatic driver torque, sudden
change of the automatic driver torque can be prevented, at the time when the magnitude
relationship between the steering wheel angle and the target steering wheel angle is
reversed, and the stability of steering control improves.
[0099]
15 In Fig. 25, when going straight ahead, the driver torque is converged to a fixed value.
And, for the same reason as in the case of clip processing, when the product of the
maximum value of the automatic driver torque and the minimum value of the gain is
smaller than the minimum value of the second threshold θ K, the convergence value
becomes a value which is equal to the second threshold θ K. In the below, explanation
20 will be made about the reason. Since it is necessary to maintain the steering wheel at 0
deg, when the vehicle is during an offset driving on a straight road, the sum of the driver
torque and the automatic driver torque is 0 Nm, when the support driver torque is 0 Nm.
For that reason, when a driver torque is converged, the gain is also converged.
[0100]

53
There are three kinds of patterns in which the gain converges. The first kind is a
pattern in which the increment of gain is positive, and the gain converges to a maximum
value, the second kind is a pattern in which the increment of gain is negative and the
gain converges to a minimum value, and the third kind is a pattern in which the
5 increment of gain is 0 and the gain converges to a suitable value. Since the gain must be
on the decrease during an override, it is considered that the candidate is the second kind
pattern or the third kind pattern. Since the increment of gain needs to be 0 in order to
converge by the third kind pattern, the driver torque needs to be equal to the second
threshold. Therefore, the convergence value of the driver torque becomes a value which
10 is equal to the predetermined second threshold.
[0101]
However, when the product of the automatic driver torque and the minimum value
of the gain is larger than the minimum value of the second threshold θ K, the automatic
driver torque converges to a value which is larger than the predetermined second
15 threshold θ K. Since a driver torque which is larger than the second threshold θ K is
needed for the cancellation of this automatic driver torque, the increment of gain
becomes negative and convergence is produced by the second kind pattern. That is, the
gain is converged to a minimum value. In order to secure the convergence by the third
kind pattern, the product of the maximum value of the automatic driver torque and the
20 minimum value of the gain needs to be smaller than the second threshold θ K.
[0102]
From the above, when the product of the maximum value of the automatic driver
torque and the minimum value of the gain is smaller than the second threshold θ K, the
convergence value of the driver torque during an offset driving becomes a value which

54
is equal to the second threshold θ K, regardless of the curvature. In this way, according
to the constitution of the Embodiment 5, the convergence value of the driver torque
during an offset driving can be predicted. Thereby, the benefit obtained here is that it
becomes easy to adjust the reaction force which a driver receives from the steering
5 wheel.
[0103]
In this way, according to the constitution of the Embodiment 5, by computing a gain
through the increment of gain, even though the driver torque changes suddenly, the gain
itself does not change suddenly, and the stability of the steering control can be improved.
10 Moreover, when computing the automatic driver torque based on the deviation of a
steering wheel angle and a target steering wheel angle, sudden change of the automatic
driver torque can be prevented, at the time when the positive or negative of the
deviation of a steering wheel angle is reversed, and the stability of the steering control
improves. Moreover, the minimum value of a gain and the minimum value of the
15 second threshold θ K are designed so that the product of the maximum value of the
automatic driver torque and the minimum value of the gain may become smaller than
the minimum value of the second threshold θ K. Thereby, during an offset driving, the
driver torque converges to the predetermined second threshold θ K, regardless of the
curvature of a driving lane. Then, the benefit obtained here is that the reaction force
20 which is applied to a driver can be made constant, regardless of the curvature.
[0104]
Embodiment 6.
In the Embodiment 5, the gain of the automatic driver torque is changed according
to the driver torque. In addition, it is allowed to have a larger second threshold θ K,

55
according to the magnitude of the degree of the first deviation. Since the driver torque
during an offset driving converges to the second threshold θ K, a larger convergence
value of the driver torque can be obtained, as the degree of the first deviation becomes
larger. That is, the reaction force which is applied to a driver can be increased, and the
5 risk of deviating from the lane during an override can be reduced.
In the below, explanation will be made about the Embodiment 6. Explanation which
overlaps with the Embodiments 1 - 5 is omitted here.
[0105]
Fig. 26 is a block diagram which shows the schematic constitution of the steering
10 control system of the Embodiment 6. Explanation about portions which are common to
Fig. 1 and Fig. 21 will be omitted.
The difference from Fig. 21 is that the torque correction computing part 201B is
equipped with a second threshold computing part 281. The second threshold computing
part 281 computes a second threshold θ K, based on the degree of a first deviation. Here,
15 the degree of the first deviation is a lateral position at a host vehicle position, for
example. Or, the degree of the first deviation is a lateral position at a look-ahead
distance, a distance to the lane marking of a host vehicle driving lane at a host vehicle
position, a distance to the lane marking of a host vehicle driving lane at a look-ahead
distance, or a time until a host vehicle crosses the lane marking of a host vehicle driving
20 lane. In addition, any variable which represents the degree of the deviation of a host
vehicle may be used.
The gain computing part 280 computes the weight based on the second threshold θ K
which is computed in the second threshold computing part 281.
[0106]

56
Fig. 27 is a flow chart which shows the procedure of a steering control device of the
Embodiment 6. Explanation about portions which are common to Fig. 5 and Fig. 22 will
be omitted.
In Step S281 of Fig. 27, a second threshold θ K is computed in the second threshold
5 computing part 281. Computation is conducted so that the second threshold θ K may
become larger, as the degree of the deviation becomes larger. For the computation of the
second threshold θ K, a map may be used, or a suitable monotonically increasing
function may be used.
[0107]
10 Next, in Step S280 of Fig. 27, a gain K is computed in the gain computing part 280.
The gain is computed in the same procedure as Step S280 of Fig. 22, except that the
second threshold θ K, which is computed in the second threshold computing part 281, is
used for the computation of the gain K.
Fig. 28 is a drawing which shows an example of the map M θK for computing the
15 second threshold θ K based on the degree of the first deviation in the second threshold
computing part 281. Here, it is assumed that the absolute value of a lateral position | e0 |
at a host vehicle position is the degree of a first deviation. This map is designed so that
the second threshold θ K may monotonically increase, according the absolute value of a
lateral position | e0 |. Since the driver torque during an offset driving converges to the
20 second threshold, a larger convergence value of the driver torque can be obtained, as the
degree of the first deviation becomes larger. That is, a large reaction force which is
applied to a driver can be obtained and the benefit obtained here is that the risk of
deviating from a lane during an override can be reduced.
[0108]

57
It is worth noticing that, as explained in the Embodiment 5, in order to make the
driver torque during an offset driving converge on the second threshold θ K, the product
of the maximum value of the automatic driver torque and the minimum value of the
gain needs to be smaller than the minimum value of the predetermined second threshold
5 θ K. Therefore, it is necessary to design a Map M θK so that the second threshold θ K may
always become larger than the maximum value of the automatic driver torque and the
minimum value of the gain.
Moreover, in the present Embodiment, the absolute value of a lateral position | e0 | is
used to the input of a map. However, it is allowed to use | e0 | which is multiplied by a
10 coefficient, where the coefficient becomes 1 at the outer side of a curve and -1 at the
inner side of a curve. As a result, the convergence value of the driver torque can be
changed at the outer side and inner side of a curve.
In this way, according to the constitution of the Embodiment 6, the second threshold
θ K changes according to the degree of the first deviation. Since the driver torque during
15 an offset driving converges to the second threshold θ K, a larger convergence value of
the driver torque can be obtained, as the degree of the first deviation becomes larger.
That is, a large reaction force which is applied to a driver is obtained, and the risk of
deviating from a lane during an override can be reduced.
[0109]
20 Embodiment 7.
In the Embodiment 1, the increment of weight is computed based on the driver
torque, and the weight and the automatic driver torque clip value are changed. In the
Embodiment 5, the increment of gain is computed based on the driver torque and the
gain of the automatic driver torque is changed. However, it is allowed to combine the

58
weight and the gain. By combining these, the convergence speed of the driver torque
during an offset driving can be accelerated. For example, in the case where only one of
the two, weight or gain, is used, and in the case where a LKAS torque clip value is very
large ( for example, 50 Nm or so ), the automatic driver torque will change greatly, even
5 when the weight or the gain changes slightly.
[0110]
For that reason, the speed at which the driver torque converges during an offset
driving to the first threshold θ W, or to the second threshold θ K becomes slower. On the
other hand, when the weight and the gain are combined, the convergence speed of the
10 driver torque can be accelerated.
In the below, explanation will be made about the Embodiment 7. Explanation which
overlaps with the Embodiments 1 - 6 is omitted here.
[0111]
Fig. 29 is a block diagram showing the schematic constitution of a steering control
15 system of the Embodiment 7. Explanation about portions which are common to Fig. 1
and Fig. 21 will be omitted. The difference from Fig. 1 and Fig. 21 is the constitution of
the torque correction computing parts 201A and 201B. In this Embodiment 7, the
steering control device is configured so that both the gain computation and the weight
computation may be conducted.
20 The gain processing part 290 multiplies an automatic driver torque by a gain, where
the automatic driver torque is clip processed in the clip processing part 260.
[0112]
Fig. 30 is a flow chart which shows the procedure of the steering control device of
the Embodiment 7. Explanation about portions which are common to Fig. 5 and Fig. 22

We Claim:
[ Claim 1 ]
A steering control device, comprising;
5 a driver support torque computing part which computes a driver support torque
according to a driver torque,
an automatic driver torque computing part which computes an automatic driver
torque according to road condition, and
an additional driver torque computing part which computes an additional driver
10 torque according to the driver support torque and the automatic driver torque,
wherein the additional driver torque computing part receives an output from at least
one of a first torque correction computing part and a second torque correction
computing part;
wherein the first torque correction computing part includes:
15 a weight computing part which computes an increment of weight based on a
magnitude of the driver torque, and accumulates the increment of weight to generate a
weight,
a clip value computing part which computes an automatic driver torque clip value
according to the weight, and
20 a clip processing part which clip processes the automatic driver torque with the
automatic driver torque clip value, to limit an upper limit value and a lower limit value
thereof, and outputs a clip processed automatic driver torque, to the additional driver
torque computing part, and
the second torque correction computing part includes:
a gain computing part which computes an increment of gain according to the driver
torque, and accumulates the increment of gain to generate a gain, and
a gain processing part which outputs an automatic driver torque multiplied by the
gain, to the additional driver torque computing part.
5 [ Claim 2 ]
The steering control device according to Claim 1, comprising the first torque correction
computing part,
wherein, when the automatic driver torque clip value becomes smaller, as the weight
is smaller,
10 the weight computing part computes an increment of weight as positive, if the
magnitude of the driver torque is smaller than a first threshold, which is determined
beforehand, and the weight computing part computes the increment of weight as
negative, if the magnitude of the driver torque is larger than the first threshold, and
when the automatic driver torque clip value becomes smaller, as the weight is larger,
15 the weight computing part computes the increment of weight as negative, if the
magnitude of the driver torque is smaller than the first threshold, and the weight
computing part computes the increment of weight as positive, if the magnitude of the
driver torque is larger than the first threshold.
[ Claim 3 ]
20 The steering control device according to Claim 2,
wherein, in the weight computing part, a minimum value of the automatic driver
torque clip value is set to be larger than a maximum value of the first threshold.
[ Claim 4 ]
The steering control device according to Claim 2,

80
wherein, in the weight computing part, a minimum value of the automatic driver
torque clip value is set to be smaller than a minimum value of the first threshold.
[ Claim 5 ]
The steering control device according to Claim 4,
5 wherein the weight computing part includes a first threshold computing part which
computes the first threshold according to a degree of a first deviation from a driving
lane of a host vehicle, and
the first threshold is set according to the degree of the first deviation.
[ Claim 6 ]
10 The steering control device according to Claim 1, comprising the second torque
correction computing part,
wherein the gain computing part computes an increment of gain according to the
driver torque, and accumulates the increment of gain to generate a gain,
where the gain computing part computes the increment of gain as negative, if the
15 magnitude of the driver torque is larger than a second threshold which is determined
beforehand, and the gain computing part computes the increment of gain as positive, if
the magnitude of the driver torque is smaller than the second threshold.
[ Claim 7 ]
The steering control device according to Claim 6, comprising the second torque
20 correction computing part,
wherein, in the gain computing part, a minimum value of the gain and a minimum
value of the second threshold is set, so that a product of a maximum value of the
automatic driver torque and a minimum value of the gain may become smaller than a
minimum value of the second threshold.

81
[ Claim 8 ]
The steering control device according to Claim 7,
wherein the gain computing part includes a second threshold computing part which
computes the second threshold according to a degree of a first deviation from a driving
5 lane of a host vehicle, and
the second threshold is set according to the degree of the first deviation.
[ Claim 9 ]
The steering control device according to Claim 1, comprising the first torque correction
computing part and the second torque correction computing part,
10 wherein the gain processing part multiplies the clip processed automatic driver
torque by the gain.
[ Claim 10 ]
The steering control device according to Claim 9,
wherein, in the gain processing part, a minimum value of the automatic driver
15 torque clip value, a minimum value of the gain, a minimum value of a first threshold,
and a minimum value of a second threshold is set, so that a product of a minimum value
of the automatic driver torque clip value and a minimum value of the gain may become
smaller than the minimum value of the first threshold and the minimum value of the
second threshold.
20 [ Claim 11 ]
The steering control device according to Claim 10, comprising a first threshold
computing part which computes the first threshold according to a degree of a first
deviation from a driving lane of a host vehicle, and a second threshold computing part
which computes the second threshold according to a degree of a first deviation from a

82
driving lane of a host vehicle,
wherein the first threshold and the second threshold are set according to the degree
of the first deviation.
[ Claim 12 ]
5 The steering control device according to any one of Claims 5, 8, and 11,
wherein a distance from a center of a driving lane of a host vehicle to the host
vehicle is set as the degree of the first deviation.
[ Claim 13 ]
The steering control device according to any one of Claims 1 to 5, and 9 to 11,
10 wherein the weight computing part changes the increment of weight according to a
degree of a second deviation from a driving lane of a host vehicle.
[ Claim 14 ]
The steering control device according to any one of Claims 1, and 6 to 11,
wherein the gain computing part changes the increment of gain according to a
15 degree of a second deviation from a driving lane of a host vehicle.
[ Claim 15 ]
The steering control device according to Claim 13,
wherein, in the weight computing part, a distance from a center of the driving lane
to a look-ahead distance, or a speed of the host vehicle in a direction perpendicular to a
20 direction of the driving lane is set as the degree of the second deviation.
[ Claim 16 ]
The steering control device according to Claim 14,
wherein, in the gain computing part, a distance from a center of the driving lane to a
look-ahead distance, or a speed of the host vehicle in a direction perpendicular to a

83
direction of the driving lane is set as the degree of the second deviation.
[ Claim 17 ]
The steering control device according to any one of Claims 1 to 5, 9 to 11, and 13,
wherein, when a direction of the automatic driver torque and a direction of the
5 driver torque are the same, or, when a direction of a difference between a target steering
wheel angle and a real steering wheel angle and a direction of the driver torque are the
same,
the weight computing part computes the increment of weight so that the automatic
driver torque clip value may increase.
10 [ Claim 18 ]
The steering control device according to any one of Claims 1, 6 to 11, 14, and 15,
wherein, when a direction of the automatic driver torque and a direction of the
driver torque are the same, or, when a direction of a difference between a target steering
wheel angle and a real steering wheel angle and a direction of the driver torque are the
15 same,
the gain computing part computes the increment of gain so that the increment of
gain may increase.
[ Claim 19 ]
The steering control device according to any one of Claims 1 to 18, further comprising a
20 curvature compensation torque computing part which computes a curvature
compensation torque, based on a curvature of a driving lane of a host vehicle and a
speed of the host vehicle, for making a steady circular turn at the speed and in addition
at the curvature,
wherein the additional driver torque is computed based on the driver support torque,
the automatic driver torque, and the curvature compensation torque.