Design of Visual Displays for Driving Simulator Research
G. John Andersen
Department of Psychology
University of California, Riverside
Vision Issues in Driving Simulators
• Driving tasks require accurate perception of depth and distance in the 3D scene– What sources of information are critical– Integration of sources– Conflict between sources
• Visual Display – Display characteristics (Aliasing, Flicker,
Update,Refresh, Resolution)
Failure to Consider these issues in Driving Simulators
• Poor task performance• Low external validity of results• Poor transfer of training• Increased risk of simulator sickness
Driving tasks and Perception of Depth and Distance
, 96
Cutting, 1996
Depth
Con
trast
Depth (meters)
Occlusion
1 10 100
Size constancy
Cast Shadows
Disparity
Motionparallax
Convergence
Aerial
Overall scene depth
Layout of Objects
Properties of Objects
Viewer-centered
Information
Accommodation
Linear Perspective
Disparity
Motion Parallax
Linear Perspective
Relative Size
Motion Parallax
Texture (perspective)
Specular Highlights
Object-centered
Information
Surrounding Frame
Compression Gradient
Structure from Motion
Compression Gradient
Structure from Motion
Texture (non-perspective)
Shading
Andersen, Braunstein, & Saidpour, 1998
Driving Tasks and Perception
• Multiple sources of information available for driving tasks– What information is critical for what tasks
• Multiple sources for a specific task• Different tasks require different information
sources
– Use of multiple sources• Cue interactions• Cue conflicts
Multiple sources for a specific task: Collision Detection
• Model based on analysis of visual information available to driver (Andersen & Sauer, 2004)– Use of 5 parameters
motion parallax; optic flow)
d/dt (motion parallax; optic flow)
d/dt
ds (texture gradients, disparity, relative size)
Front View Top View
t=1
t=0
specifies the time to contact during constant velocitycollisions
a is the bearing of an object relative to a reference
ddt, dsd/dt, d/dt,
ddt, ds
ddt, ds
d/dt, d/dt, ddt, ds
ddtds ddt, ds
ddt, ds
d/dt, ds
ddt
d/dt, ddt, ds
d/dt, d/dt, d/dt, ds ddt
d/dt
d/dt, ddt, ds
d/dt, ddt, ds
d/dt, ddt, ds
d/dt, ddt, ds
d/dt, ddt, ds
d/dt, ddt, ds
Vehicle Motion
ObjectMotion
F = Fixed Speed V = Variable SpeedS = Straight Path C = Curved Path
No
F/S
V/S
F/C
V/C
No F/S V/S F/C V/C
Is =0? Or d/dt0?(constant angular
direction)
Is >0?(expansion)?
Yes
No
Yes
Is d/dt =0?(Constant expansion)
No
Derive
d/dt
Is d/dt >=-0.5?
No
Collision Object
Derivedv-ds
Is dv-ds >0?
Yes
Non-collision Object
Derive
Derive d/dt(angular direction)
Is dv-ds =0?(Constant expansion)
No
Yes
SelectObject
Different tasks and different sources of information
Two tasks with different sources of information
- Collision Detection (constant speeds and linear trajectories): Optical Expansion (TTC) and
-Vehicle Speed: Edge Rate
Question: Does information from one task affect use of information for second task
Experiment• Issue: Does information for one
task affect use of information for a different task
• Task: collision detection
• Independent Variables– Collision or non-collision object
– Time to contact (TTC)
– Different combinations of vehicle speed and object speed
• Dependent variable– Sensitivity (d’)
Vehicle / Object Velocity
d'
2.4
2.6
2.8
3.0
3.2
3.4
3.6
3.8
4.0
4.2
4.4
4.6
0/-0.8 0.2/-0.6 0.4/-0.4 0.2/-0.6
TTC 3.5 s
TTC 4.5 s
TTC 5.5 s
Simulators depict a 3D world
Perception of 3D scene includes-Texture gradients-Motion parallax-Occlusion-other information sources
Visual system also has cues for perceiving the distance to the display
3 Information sources for depth and distance of Display
- Binocular Disparity
- Eye convergence
- Accommodation
, 96
Cutting, 1996
Depth
Con
trast
Depth (meters)
Occlusion
1 10 100
Size constancy
Cast Shadows
Disparity
Motionparallax
Convergence
AerialAccommodation
Visual Cue Conflicts between display cues and 3D scene cues
Visual Cue conflict:
- Disparity, convergence and accommodation information specify display is flat
- Motion parallax, texture gradients, and other cues specify scene is 3D
Human Factors Solutions to these Issues
• Determine perceptual information needed for driving tasks of interest
• Design displays to optimize performance of tasks– Consider interactions of different information sources
• Minimize cue conflicts• Design for optimal performance not for high
fidelity– Question need of high fidelity rendered scenes, control
dynamics and fixed/variable base simulator