Andersen

Vehicle Characteristics and Car
Following
George J. Andersen
Department of Psychology
University of California, Riverside
Funded by NIH AG13419-06 PATH Project MOU 4220
Perceptual Tasks in Driving
•
•
•
•
Collision Detection
Obstacle avoidance
Longitudinal control (car following)
Lateral control (steering)
Perceptual Tasks in Driving
•
•
•
•
Collision Detection
Obstacle avoidance
Longitudinal control (car following)
Lateral control (steering)
Driving is a skill dependent on visual information
Use of simulators requires accurate presentation
of visual information used by drivers
Complexity of Collision Detection:
Event Specification
• Vehicle motion
– Speed
• Constant or varying (accelerating/decelerating)
– Path
• Straight or curved
• Object motion
– Speed
• Constant or varying (accelerating/decelerating)
– Path
• Straight or curved
Complexity of Collision
Detection:
• Model (Andersen & Sauer, 2004) based on
analysis of visual information available to
driver
– Use of 5 parameters
t
dt/dt
a
da/dt
ddiff
Front View
Top View
t=0
t=1
t=q/Dq
q
t specifies the time to contact during constant velocity
collisions
0.4
-.9
0.3
=dt/dt
-.6
Deceleration(g)
t
-.7
-.55
0.2
-.5
-.45
0.1
-.3-.4
-.1
300
200
100
1
0.0
Frames
dt/dt used during deceleration (braking control)
When dt/dt = -0.5 vehicle will reach zero velocity at obstacle
a is the position of
object in visual field
When a = 0 object is on a collision path
Useful when path of motion is linear
da/dt is the change in
position of object in
visual field
When da/dt = k object is on a collision path
Useful when path of motion is curvilinear
ddiff is comparison of two distance estimates:
dv – distance vehicle will traverse before reaching zero velocity
ds – distance of collision object
ddiff = dv – ds
dv = 1.5v2/a
v = edge rate (number of texture elements that
pass position in visual field)
a = change in number of texture
Elements that pass position in visual field
ds = (s)tan-1 q
s = size of object
q = visual angle of object
Size information and safe deceleration to a stop
% of total time regulated
20
18
Stop Sign Size
16
200 units
400 units
14
12
10
8
6
4
2
0
-.92 -.81 -.70 -.59 -.49 -.38 -.27 -.16 -.05
taudot
.05
% of total time regulated
Edge rate information and safe deceleration to a stop
18
16
14
12
10
8
6
4
2
0
Texture edges / sec
7.2
72
-.92
-.70
-.81
-.49
-.59
-.27
-.38
BINS
-0.05
-.163
.05
Vehicle Motion
No
F/S
V/S
F/C
da/dt, t
V/C
No
t
a
dt/dt, ds
t
F/S a
t
a
dt/dt, ds
dt/dt, ds
dt/dt, ds
da/dt, ds
da/dt, t
a
a
dt/dt
dt/dt, ds
da/dt, t
da/dt, ds
da/dt, t
dt/dt
da/dt
t
Object V/S dt/dt, ds
Motion
F/C
V/C
da/dt, t
da/dt, t
dt/dt, ds
da/dt, t
a
da/dt, t
dt/dt, ds
dt/dt, ds
da/dt, t
da/dt, t
da/dt, t
da/dt, t
da/dt, t
dt/dt, ds
dt/dt, ds
dt/dt, ds
dt/dt, ds
dt/dt, ds
F = Fixed Speed V = Variable Speed
S = Straight Path C = Curved Path
Optical Information for Car
Following
• Information for specifying distance and
change in distance
• Information for specifying speed and
change in speed
Top View
t=0
t=1
a
Front View
Da associated with change in distance
due to change in speed
Parameters of Car Following
Model
a’
– Initial visual angle of lead vehicle
a
– Current visual angle
da/dt
– Instantaneous change in visual angle
J, k
– Weighting scalar constants
(km/hr2)
Acceleration
acceleration
acceleration
j 
1
a

  k da

dt
a' 
1
2m
w


a' 2 atan 

.001

timegap
 FVv 
desired_a( v ) = 2
atan

v
Lead Car
desired_timegap = 1.1


 desired_timegap 

3600 

0.3
0.2
desired_a ( velocity
a’ )
0.1
Distance headway
α
0
0
50
100
velocity
Desired time gap = 1.1s
W = width of lead vehicle
Driver
FVv = following vehicle
(driver) speed
150
Human Factors Experiments
• Maintain distance behind lead vehicle that
varied speed
- sine function
- ramp function
- sum of sines
function
Frequency: 0.0513
Amplitude: 15 k/h
65
Velocity (k/h)
55
45
35
25
15
0
200
100
400
300
600
500
800
700
TIME (20 Frames / Sec)
1000
900
1200
1100
140
006
130
120
110
data Fcvel
100
data Lcvel
model Fcvel
90
80
70
60
1
97
193 289 385 481 577 673 769 865 961 1057 1153 1249 1345 1441 1537 1633
0.09
5
0.08
4.5
0.07
4
0.06
3.5
0.05
data range
model range
0.04
data timegap
model timegap
3
2.5
2
1.5
0.03
1
0.02
0.5
0.01
0
0
1
127 253 379 505 631 757 883 1009 1135 1261 1387 1513 1639
1
205 409 613
817 1021 1225 1429 1633
Do drivers use visual information other than visual
angle?
Edge Rate Information:
Used for Perceived
Driver (following vehicle)
speed
Edge Rate and Collision Detection: Moving Objects
t=
a
da/dt
Edge Rate and Collision detection during braking:
Static objects
t
=dt/dt
Car Following and edge rate Experiment
Task: Car following to sine
wave function
Independent Variables:
Presence or absence of scene
Frequency and amplitude of
lead vehicle speed
Prediction:
If edge rate used then more
accurate tracking performance
when scene present as
compared to scene absent
2.2
2.0
Scene absent
Scene Present
Control Gain
1.8
1.6
1.4
1.2
1.0
0.8
0.6
5 kph
15 kph
25 kph
5 kph
FREQ: 0.513 Hz
15 kph
25 kph
FREQ: 0.1111 Hz
-10
-20
Phase Angle
-30
Scene Absent
Scene Present
-40
-50
-60
-70
-80
5 kph
15 kph
FREQ: 0.0513 Hz
25 kph
5 kph
15 kph
FREQ: 0.1111 Hz
25 kph
Ongoing Research: Car Following in
Traffic
Edge Rate and Moving Vehicles
Dual task performance
car following
Detect Light Change
Edge Rate Information
Presence of other moving
vehicles
Edge Rate and Reduced Visibility
Dual task performance
car following
Detect Light Change
Edge Rate Information
Presence of Fog
Simulation Design Issues and
Recommendations
• Simulation displays should be designed to
optimize use of visual information
– Understanding how best to do this requires
understanding what are the sources of
information
Simulation Design Issues and
Recommendations
• Factors that directly affect availability of
information sources
– Display characteristics (e.g., frame rate, spatial
resolution, monitor update and flicker)
– 3D model characteristics (e.g., complexity of
world model, lighting, and texturing)
– Viewing characteristics
• (e.g., conflicting accommodation, eye vergence)
• Viewing from design eye of simulation