on Primary and

duewestseaurchinAI and Robotics

Nov 14, 2013 (3 years and 8 months ago)

73 views

Influence of
Spatial Ability
on Primary and
Secondary
Space
Telerobotics
Operator
Performance
Zakiya Tomlinson, Joseph Silverman,
Charles Oman, Andrew Liu, Alan Natapoff
MVL Space
Teleoperation
Research
December 10, 2008
Outline

Background

Telerobotics Training

Primary/Secondary Operators

Spatial Ability

Experiment Design

Hypotheses

Virtual Environment

Experiment 1 Overview

Experiment 2 Methods

Results

Primary Operator

Secondary Operator

Gaze Analysis
Photo Credit: NASA
MVL Space
Teleoperation
Research
December 10, 2008
Telerobotics Training

Teleoperation

A form of telerobotics where the human operator can
directly monitor/control the remote device

NASA Generic Robotics Training (GRT)

Basic instruction for all astronauts

Classroom lessons

Desktop simulation of a 6 Degree of Freedom Arm (BORIS)

Controllable Mock
-
up of the Space Station arm (MRMDF)

Takes approx 30 hours, not including practice time

Customizing the training for each student could make
learning easier and more efficient
MVL Space
Teleoperation
Research
December 10, 2008
Telerobotics Training
MRMDF Training Facility
ISS Robotic Workstation Trainer
MVL Space
Teleoperation
Research
December 10, 2008
Primary and Secondary Operators

Astronauts work in
pairs when controlling
a robotic arm in space

Primary Operator:
manipulates the
controls

Secondary Operator:
assists with changing
camera views and
setting up the arm,
and monitors the task
Photo Credit: NASA
MVL Space
Teleoperation
Research
December 10, 2008
Spatial Ability

Spatial abilities determine how people remember,
visualize, or transform spatial information.

Most relevant areas to telerobotics:

Spatial Visualization: ability to manipulate an object into a
new arrangement

Spatial Orientation: ability to imagine how an object will
look after its has been rotated

Mental Rotation (MR): ability to mentally rotate objects

Perspective Taking (PT): ability to visualize objects from a new
perspective within an environment

Three tests used:

MRT
,
PSVT
(3
-
D PT) and PTA (2
-
D PT)
MVL Space
Teleoperation
Research
December 10, 2008
Hypotheses

Subjects with higher spatial ability skills would:

Perform better at primary operator tasks

Trial time, movement fluidity, error from shortest path

Perform better at secondary operator tasks

Correct and timely problem detection

Perform better with a large disparity between camera
-
and control
-
frame orientations

Subjects with poor spatial orientation and spatial
visualization skills would fixate on a single view
instead of spreading their attention between all
of the monitors.
MVL Space
Teleoperation
Research
December 10, 2008
Camera 2
Camera 1
Camera 3
Camera 4
End
-
Effector
Camera
Virtual Environment

Modeled after the NASA
BORIS training tool

5 possible camera views

4 fixed cameras (one in each
corner)

1 mobile camera on the arm’s
end
-
effector

Disparity

orientation of arm’s base
(control
-
frame) vs. orientation
of the camera (camera
-
frame)

subjects trained under both
high and low disparity
conditions
MVL Space
Teleoperation
Research
December 10, 2008
Experiment 1 Overview
Task Design:

Manoeuvre the arm to the target box

Position the end
-
effector 1.5m above
the top surface of the box

Align with the grapple target

Results

Better (predictor) 3
-
D
Perspective Taking
Scores = smaller error
from best path

Low scores = weaker
at judging clearances

High disparity between
control
-
and camera
-
frame = more
movements and larger
error from best path
MVL Space
Teleoperation
Research
December 10, 2008
Experiment 2 Methods
Translational
Hand Controller
Rotational
Hand Controller

6 Primary Operator Trials

Move end
-
effector to
point specified on maps
MVL Space
Teleoperation
Research
December 10, 2008
Experiment 2 Methods
Camera 2
Camera 3
Exmaple
shows X/Y
plane
position
using
cameras 2
and 3

Subjects were
to work as
quickly as
possible while
avoiding
problems such
as collisions

A camera
recorded where
subjects were
looking while
they worked
MVL Space
Teleoperation
Research
December 10, 2008
Experiment 2 Methods

Space Station Virtual
Environment

Modeled after ISS as
of Fall 2007 (STS
-
120)

32 trials

Subjects observed
playback of robotic
operations and
stopped the arm if
they saw a problem

Singularities

Clearance Violations

Unexpected Motions
MVL Space
Teleoperation
Research
December 10, 2008
Experiment 2 Methods

Singularities

Clearance Violations

Unexpected Motions
“Now I will move aft over the truss.”
MVL Space
Teleoperation
Research
December 10, 2008
Primary Operator Results

Spatial Ability Scores

No statistical difference from
Astronaut test group (n = 40)

No statistical differences
between returning subjects
(n = 9) and naïve (n = 11)

Female subjects scored
significantly lower than males

Manual Control Ability

Simultaneously use both
hand controllers to trace the
end
-
effector around a shape

Returning subjects had
higher scores (p = 0.011)
Two-sample t-test
1
0
REPEAT
0
4
8
12
16
Count
-10
0
10
20
30
40
BMC
0
4
8
12
16
Count
Two-sample t-test
1
0
REPEAT
0
4
8
12
Count
-10
0
10
20
30
40
BMC_ORIG
0
4
8
12
Count
Male
Mean (SD)
Female
Mean (SD)
MRT
24.64 (7.56)
8.60 (4.78)
PSVT
18.00 (5.25)
9.20 (3.70)
PTA
22.50 (3.42)
16.08 (4.33)
MVL Space
Teleoperation
Research
December 10, 2008
Primary Operator Results

Learning Effects

Trial completion time (p = 0.003)

Fluidity of movements (p < 0.001, p = 0.001)

Percentage of time spent moving (p < 0.001)

Number of direction changes (p = 0.005)
MVL Space
Teleoperation
Research
December 10, 2008
Primary Operator Results

Effect of PTA Score

Percentage of time spent moving (p = 0.008)

Error from the final target position (p = 0.001)

Fluidity of movements (p < 0.001, p < 0.001)
MVL Space
Teleoperation
Research
December 10, 2008
Primary Operator Results

Effect of Disparity

Lower errors from the
best path under the
low disparity condition
(p = 0.008)

Below average PTA
scorers took longer to
complete trials under
high disparity
(p < 0.001)

Effect of Learning Strategy

Subjects who reported carefully studying the environment had:

Shorter trial completion times (p = 0.001)

Better movement fluidity (p = 0.024, p = 0.003)

Higher percentage of time spent moving (p = 0.004)
MVL Space
Teleoperation
Research
December 10, 2008
Secondary Operator Results

Learning Effects

More correct detections (p = 0.001)

Fewer false alarms (p = 0.005) and missed detections
(p < 0.001)
0
1
2
3
4
LESSON
0
1
2
3
4
5
6
7
CORRECT DETECTIONS
0
1
2
3
4
LESSON
0
1
2
3
4
5
6
FALSE ALARMS
0
1
2
3
4
LESSON
0
1
2
3
4
5
6
MISSES
MVL Space
Teleoperation
Research
December 10, 2008
Secondary Operator Results

Perspective Taking

Above average PSVT
scorers had higher
total (weighted)
scores (p = 0.043)

Above average PTA
scorers had more
timely detections
(p = 0.040)
Two-sample t-test
1
0
PAY_GROUP
0
1
2
3
4
5
6
7
Count
0
10
20
30
PSVT
0
1
2
3
4
5
6
7
Count
MVL Space
Teleoperation
Research
December 10, 2008
Gaze Analysis
-
Thinking in 3
-
D

Maps:

Top View

Side View
Each of the
3 monitors
gives a
unique
perspective
While controlling the robotic arm, subjects must choose what is the best way
to look at this information
MVL Space
Teleoperation
Research
December 10, 2008
Looking at right screen
Looking at left screen
Looking at middle screen
Looking at Map
Where Subjects Can Look
MVL Space
Teleoperation
Research
December 10, 2008
Gaze Analysis Results

Probability of Switching between Options

Left

Middle = Right

Middle

Same probability for Maps

any monitor

Future Work

Does matrix vary between high and low scorers on spatial
ability tests?
Left
Middle
Right
Maps
Left

0.545
0.065
0.389
Middle

0.340
0.424
0.236
Right

0.089
0.547
0.364
Maps

0.329
0.344
0.327
MVL Space
Teleoperation
Research
December 10, 2008
Gaze Analysis Results

Subjects who switched where the were looking more
frequently had:

High (above 21) BMC scores (p = 0.002)

Above average MRT scores (p = 0.037)

Above average PTA scores (p = 0.012)
MVL Space
Teleoperation
Research
December 10, 2008

Below average PSVT scorers spent a larger percentage
of their time looking at the map (p = 0.007)
Gaze Analysis Results
MVL Space
Teleoperation
Research
December 10, 2008
Gaze Analysis Results

No significant effect found between operator performance and gaze

Future studies?
MVL Space
Teleoperation
Research
December 10, 2008
Conclusions

Results verified hypotheses:

Performance on telerobotic tasks is affected by spatial abilities

High camera
-
/control
-
frame disparities negatively affect
telerobotic performance

Perspective taking ability affects performance under high
disparity conditions

The tendency to fixate on a single screen during a telerobotic
task is affected by spatial abilities

Overall performance as a secondary operator and timely
problem detection are affected by perspective taking ability

Experiments simulate very early robotics training
activities at NASA

How does early performance relate to final performance?
MVL Space
Teleoperation
Research
December 10, 2008
Questions?
Acknowledgements
Dr. Charles Oman
Dr. Andrew Liu
Dr. Alan Natapoff
NASA JSC Robotics Instructors
Daniel Burbank, USCG
NSBRI
Influence of
Spatial Ability
on Space
Teleoperator
Camera
Selection
Teresa Pontillo, Zakiya Tomlinson
Charles Oman, Andrew Liu, Alan Natapoff
MVL Space
Teleoperation
Research
December 10, 2008
Outline

Objective

Virtual Environment

Methods

Pilot Experiment
Photo Credit: NASA
MVL Space
Teleoperation
Research
December 10, 2008
Experiment Objective

To determine how a person’s individual spatial
skills influence their performance when selecting
camera views for performing space
teleoperation tasks

Hypothesis:
Subjects with higher spatial ability
scores will perform tasks more quickly and
accurately
MVL Space
Teleoperation
Research
December 10, 2008
Virtual Environment
Robotic arm
Grapple target
Modeled after NASA
BORIS environment
MVL Space
Teleoperation
Research
December 10, 2008
Viewpoints

There are 5 possible views:

Four views from numbered cameras

Window on the forward wall
Camera 2
Camera 1
Camera 3
Camera 4
Forward
Window
Subject presented
with several
different arm and
target scenarios
and must correctly
select the best
camera views
Target
MVL Space
Teleoperation
Research
December 10, 2008
Viewpoints

Learning how to properly select camera views is a big
part of the astronauts’ robotics training

For every task, three types of views are needed
Clearance
MVL Space
Teleoperation
Research
December 10, 2008
Viewpoints
Big picture view
Shows as much of the entire task as possible with a single view
BAD CHOICE
Where is the arm??
GOOD CHOICE
Situational awareness of environment
Target
MVL Space
Teleoperation
Research
December 10, 2008
Viewpoints
Clearance view
Used to determine the distance between the arm and an obstacle
BAD CHOICE
GOOD CHOICE
MVL Space
Teleoperation
Research
December 10, 2008
Viewpoints
Task view

Used to determine the arm’s distance from target during alignment

This view should be orthogonal to the top surface of the target
GOOD CHOICE
During the
final part of
alignment,
how far is the
arm from the
target?
BAD CHOICE
MVL Space
Teleoperation
Research
December 10, 2008

Procedure

Spatial Ability tests (
MRT
,
PSVT
, PTA)

12 different scenarios (3
lessons, 4 trials each) that
differ by arm and target
position

Paper maps to select
initial big picture,
clearance, and task views

Check initial selections
and make changes as
necessary
Experiment Method
MVL Space
Teleoperation
Research
December 10, 2008
Experiment Method

Principal Measurements

Spatial Ability scores

Overall performance score (number of correct selections, time)
MVL Space
Teleoperation
Research
December 10, 2008
Pilot Results

Performance vs. PTA
score

relationship between
spatial ability and overall
performance
MVL Space
Teleoperation
Research
December 10, 2008
Pilot Results

Task Time vs. Lesson
Number

Learning Effect

Great variability between
subjects
MVL Space
Teleoperation
Research
December 10, 2008
Questions?