US 11,770,270 B2 · Integrated automated driving system for maritime autonomous surface ship (MASS)

Abstract

The present invention discloses an integrated automated driving system for a maritime autonomous surface ship (MASS). The integrated automated driving system for a MASS includes a perception module for perceiving navigational environment of a MASS and obtaining real-time dynamic information of a navigation channel, hydrology, a state of the MASS and traffic environment; a communication system for transmitting data and instructions between the MASS and a shore base as well as between system modules; a data processing module for processing information obtained by the perception module; a decision-making module for identifying a current operating status of the MASS and environment according to data outputted by the data processing module, selecting actions to be taken, and generating operating instructions corresponding to the action; and an execution module for receiving operating instructions sent from the decision-making module and controlling a propeller and a rudder of the MASS through a proportional-integral-derivative (PID) controller.

An integrated automated driving system for a maritime autonomous surface ship (MASS), comprising:

a perception device configured to perceive a navigational environment of the MASS and obtain real-time dynamic information of a navigation channel, hydrology, a state or a traffic environment of the MASS;
a communication device coupled to the perception device;
a processor configured to:
process information obtained by the perception device;
identify a current operating status of the MASS and environment according to the processed information
select actions to be taken,
generate operating instructions corresponding to the action based on a deep deterministic policy gradient (DDPG) algorithm, wherein a reward function in the DDPG algorithm comprises a path tracing reward function and a collision avoidance reward function, and
control a propeller and a rudder of the MASS through a proportional-integral-derivative (PID) controller based on the operating instructions to change a state of motion of the MASS;
wherein the path tracing reward function is:
r
t
=
cxp
⁡
(
-
k
E
⁢
(
x
-
x
⁡
(
θ
t
)
)
2
+
(
y
-
y
⁡
(
θ
t
)
)
2
-
k
D
⁢
φ
-
φ
⁡
(
θ
t
)
)
2
)
+
k
V
⁢
1
(
μ
-
μ
⁡
(
θ
t
)
)
+
k
V
⁢
2
(
v
-
v
⁡
(
θ
t
)
)
+
k
M
⁢
θ
Θ
;
wherein, kE is a path deviation coefficient, kD is a course angle deviation coefficient, kV1 and kV2 are velocity coefficients, kM is a progress coefficient, which can be adjusted according to a training effect;
x and y are respectively a horizontal coordinate and a longitudinal coordinate in vessel coordinates;
θt is path tracing completion progress at a time t, Θ is a total path length, and
θ
Θ
is a ratio of a distance between a position of the MASS and a starting point to a distance between the starting point and a finishing point;
φ is the course angle of the MASS;
x(θt) y(θt) and φ(θt) are respectively an expected horizontal coordinate, an expected longitudinal coordinate and the course angle, at the time t;
u and v are respectively velocity of the MASS in a direction of surging and velocity of the MASS in a direction of swaying;
and u(θt) and v(θt) are respectively an expected value of the velocity of the MASS in the direction of surging and an expected value of the velocity of the MASS in the direction of swaying;
a computation formula of a latest collision avoidance action opportunity DLMA determined by the close quarter situation of the MASS is:
D
LMA
=
A
d
×
1
+
1
k
2
-
2
⁢
cos
⁢
α
k
wherein Ad represents a steering advance distance, K represents a velocity ratio, k=VT/V0, V0 represents current velocity of the MASS, VT represents velocity of an obstacle, if VT is 0, DLMA=Ad, and α is a relative bearing of the moving obstacle relative to the MASS;
the collision avoidance reward function is:
r
t
=
(
∑
t
=
1
n
ln
⁡
(
DCPA
t
D
s
)
2
+
k
CRI
×
CRI
t
)
+
cxp
⁡
(
-
a
1
(
V
)
D
LMA
×
∑
t
=
1
4
(
d
1
_
-
d
t
-
1
_
)
2
-
(
V
0
-
V
S
)
2
+
a
t
(
V
)
)
+
1
/
(
1
+
cxp
(
-
k
action
⁢
δ
t
×
(
a
t
(
C
)
-
a
t
=
1
(
C
)
)
)
)
wherein, DCPA represents a distance of a closest point of approach between every two collision avoidance substances;
Ds represents the closest distance for avoiding collision;
kCRI∈[0,1] represents a collision influence coefficient;
CRI represents a MASS collision probability;
C represents a course angle;
dt represents a mean distance of the obstacle at a time t;
δ represents a rudder angle of the MASS;
kaction∈[0,1] represents an action influence coefficient;
V0 and VS respectively represent current velocity and expected velocity of the MASS;
at (C) represents course variation in a decision-making set, and at (V) represents navigational velocity variation in the decision-making set.

235 more elements

Integrated automated driving system for maritime autonomous surface ship (MASS)

Life of this patent

Integrated automated driving system for maritime autonomous surface ship (MASS)

Life of this patent

Abstract

Claims11