Practical steering control for an autonomous docking based - SIP-adus

Practical steering control for an
autonomous docking based
Akihiro YAMANASHI
Advancsd Smart Mobility Co., Ltd
1
Outline
• Background
• Motivation and Purpose
• Steering control
• Experiments
• Conclusions
2
Next Generation Urban Transportation System Concept
- ART: Advanced Rapid Transit Japanese government initiated a research project on automated driving systems under CrossMinisterial Strategic Innovation Promotion Program (SIP).
ART will expand from the traditional limited operations to provide more transportation convenient
service in advanced road transportation systems. Especially, automation of a high-precision docking
to bus stop is required.
3
Experimental step and gap limit evaluation to establish
wheelchair friendly conditions
Research subjects: 12-persons (6 male and 6 female)
Age-range:
20-60 years
Experimental verdict: Step value:3cm Gap value:6cm
Estimate index
MANUAL
WHEELCHAIR
EXPERIMENT
IMAGE
ELECTRIC
WHEELCHAIR
○
Easy access
△
Requires some effort (reasonable)
▼
Outer limit
X
Assistance required
1.The allowable step limit
MANUAL WHEELCHAIR
Step value
6mm
18mm
30mm
42mm
A
○
○
△
X
B
○
○
△
▼
C
○
○
○
▼
D
○
○
○
▼
E
○
○
X
X
Evaluation
F
○
○
△
▼
○
○
△
X
Step value
6mm
18mm
30mm
42mm
ELECTRIC WHEELCHAIR
A
B
C
D
E
F
○
○
△
X
○
○
○
▼
○
○
○
▼
○
○
○
▼
○
△
○
X
○
○
○
▼
Evaluation
○
○
○
X
2.The allowable gap limit
Gap value
30mm
45mm
60mm
75mm
MANUAL WHEELCHAIR
A
B
C
D
E
F
○
○
○
▼
○
○
○
○
○
○
○
○
○
○
○
▼
○
○
○
▼
○
○
○
△
Evaluation
Gap value
○
○
○
▼
30mm
45mm
60mm
75mm
ELECTRIC WHEELCHAIR
A
B
C
D
E
F
○
○
△※
X※
○
○
○
○
○
○
○
○
○
○
○
▼
○
○
○
▼
○
○
○
▼
※front wheels to fall into the gap
Evaluation
○
○
○
▼
4
Control System for Autonomous Docking
c   K 2 e2 vr  K 3 sin e3
vr  v cos e3
Target trajectory
(Geometrically
programed coordinates)
Control Model
Steering
motors
Counter-Steering
feedforward
controller
Total distance
(Starting point)
Inclination of target
trajectory
Deviation of target
trajectory
+
+
Feedback
control of
Steering
ー
+
ー
Target
trajectory
Target deviation
(0cm)
+
Lane Keeping for Automatic Steering
• In general, a camera foresees the lane
marker .
• There are some driver model to
control the vehicle position.
• It is difficult to see a marker in
various situations: weather cond.,
close space.
• Another way is to see the lane marker
in just side of the vehicle.
6
Control with Road white line
Camera position
1.Near the Vehicle rear
door
Position of Guided white line
Guided white line
(the virtual reference trajectory)
merit and demerit
○Merit
・Performance improvement
near the rear door
○Demerit
・draw a white line at the bus stop
2.Near the Vehicle tip
center
○Merit
・Feedback control with white line is
always possible
○Demerit
・Performance drops near the rear
door
・Mixed with ordinary road white
lines
3.Near the front left end
of the vehicle
○Merit
・Feedback control with white line is
always possible
○Demerit
・Performance drops near the rear
door
Since it is difficult to realize No.1,2, it is evaluated at No3
Test Bus and System Architecture
Cameras
Cameras
Wheel speed sensors
EBS
Yaw rate sensors
Image
Processing
by PC
Steering motors
dSPACE
Autobox
8
Experiments
• The operations of acceleration and deceleration are performed
manually by a driver
• Only the steering is controlled
• The velocity of the truck is from 20 or 30 km/h to 0 km/h
• The controlled bus follows to the reference trajectory and the
reference heading angle determined from the results of manual
operation.
• The reference trajectories are three types
Type 1
Type 2
Type 3
Reference trajectory Type1
●The error to target is ±0.02m
Front door
Front door
error to target [m]
0.05
-0.01
0.02
0.03
0
1
-0.002
2
3
-0.009
-0.008
4
平均
5
-0.03
-0.02
-0.02
-0.002
-0.015
error to target [m]
1回目
0.06
0.04
0.02
0
-0.02
-0.04
-0.06
-0.01
-0.02
-0.03
2
3
-0.007
4
5
平均
-0.008
-0.015
-0.015
-0.018
-0.027
error to target [m]
Rear door
0.06
0.04
0.02
0
-0.02
-0.04
-0.06
2回目
3回目
4回目
5回目
平均
Central door
0.02
0.015
0.01
0.003
-0.005
-0.025
1回目
1
-0.02
-0.05
-0.03
0
0.005
-0.01
-0.0098
-0.01
0.00
0.01
2回目
3回目
4回目
5回目
平均
-0.005
-0.002
5回目
平均10
Rear door
0.01
0.015
0.01
-0.04
1回目
2回目
3回目
4回目
Reference trajectory Type2
●The error to target is ±0.02m
Front door
Front door
0.051
0.036
0.04
0.021
0.021
0.02
0.025
0
1
-0.004
2
3
4
5
error to target [m]
0.06
平均
0.06
0.04
0.02
0
-0.02
-0.04
-0.06
-0.04
2回目
-0.02
3回目
4回目
5回目
-0.014
平均
error to target [m]
Central door
Rear door0.036
0.001
0.02
0.014
0.0078
-0.004
1
2
3
4
0.06
0.04
0.02
0
-0.02
-0.04
-0.06
0.00
5
-0.02
-0.024
平均
0.06
0.04
0.02
0
-0.02
-0.04
-0.06
0.00
0.01
0.005
0.00
4回目
5回目
平均
0.01
0.01
-0.015
1回目
error to target [m]
0.04
-0.04
-0.005
-0.01
1回目
-0.02
0
0.005
2回目
3回目
Rear door
0.00
-0.01
1回目
-0.002
-0.02
2回目
3回目
4回目
5回目
平均
11
Reference trajectory Type3
●The error to target is ±0.02m
Front door
0.05
0.002
0.015
-0.02
2
-0.011
-0.022
0
1
-0.02
Front door
0
3
4
5
-0.009
平均
-0.0108
-0.028
-0.021
-0.024
5回目
平均
-0.04
-0.06
-0.04
-0.05
-0.04
1回目
2回目
3回目
4回目
Central door
0.06
0.04
0.02
0
-0.02
-0.04
-0.06
Rear door
0.025
0.03
0.01
0.02
-0.02
2回目
0.01
0.04
3回目
0.042
0.037
2
3
4
-0.008
5
平均
4回目
0.049
0.010
5回目
平均
0.044
0.038
0.018
0.02
1
0.014
Rear door
0.06
0.0164
0.02
-0.002
1回目
0
-0.01
0.005
0.035
0.04
0.02
0.014
0
1回目
2回目
3回目
4回目
5回目
平均
12
Experimental Results
Stopping Distance [cm]
Type of vehicle
Bus
Reference
trajectories
Type 1
Type 2
Type 3
Articulated bus
Door position
Maximal value
plus direction/minus direction
Average
(N=5)
Maximal value
plus direction/minus direction
Front door
0/-2.0
-1.0
-1.1 / -4.0
-2.4
+2.0 / -0.2
1.0
Central door
Average
(N=5)
Rear door
0/-2.7
-1.5
+4.9 / +1.8
3.8
Front door
+5.1/-0.5
2.5
+0.5 / -4.0
-1.4
+1.0 / -1.5
0.0
Central door
Rear door
+3.6/-2.4
0.8
+1.0 / -2.0
-0.2
Front door
+1.5/-4.0
-1.1
+2.0 / -2.0
-0.02
+2.0 / -2.5
0.03
+1.5 / -4.0
-0.02
Central door
Rear door
+3.5/-0.8
1.6
13
Conclusions
• As experimental performance evaluation on autonomous precision docking,
the method of the autonomous precision docking based on path following
control was described.
• The effectiveness of the proposed method was evaluated by the experimental
results.
• We confirm that the controlled truck docked with precision within about
±0.05m in the virtual reference trajectories of three types.
Thank you for your
attention!
Advancsd Smart Mobility Co., Ltd
Akihiro YAMANASHI
E-mail : [email protected]