MULTIAGENT TEAM PERFORMANCE

1
MULTI-AGENT TEAM PERFORMANCE DURING
EXTRAVEHICULAR CONSTRUCTION TASKS Col
onel Nancy J. Currie, Ph.D. NASA Johnson Space
Center
2
Background

• Purpose of the evaluation
• Space construction technology integration and
  demonstration aligned with the objectives of
  NASAs Space Architecture Team
• Evaluation of human-robot teaming strategies in
  the context of a simplified EVA assembly task
• Team performance studied to identify the
  strengths and weaknesses of each teaming
  configuration and to recommend appropriate
  division of labor
• Shared control approach developed to exploit
  complementary strengths of the EVA astronauts,
  human teleoperators, and dexterous robots
• Technology elements used in test
• Robonaut
• Advanced pressurized spacesuit (I-Suit)
• Experimental truss hardware

3
Technology Elements

• Langley Truss
• assembles quickly
• strong, stiff, light structure
• EVA-qualified

Advanced Space Architectures

• Robonaut
• humanlike dexterity
• works side-by-side with humans
• EVA-friendly

• I-Suit
• high FOV
• high mobility
• low weight (80 lbs)

4
Test Configuration
5
Multi-Agent Timeline
6
Team Performance During Truss Construction
7
Information Exchange in Human-Robot Teams
8
Robonaut Control System Architecture
9
Force Accommodation Control Laws
Force accommodation control laws are effective
safety controls and can also be effective in
multi-agent tasks where the robot follows a
teammate's lead by moving to minimize loads
When attempting to place a peg into a hole, the
impedance control law can be stiff in the
direction of insertion and compliant in
off-axes Allows the manipulator to apply forces
in the insertion direction without accumulating
additional undesired/unsafe forces
10
Topics Investigated

• Collaboration modalities between EVA and remote
  humans
• Humans interact with Robonaut in 3 roles -
  teleoperator, monitor, co-worker
• Interaction takes different forms depending on
  the configuration of the human-robot team and
  degree of robot autonomy
• Robonaut uses force and tactile sensors to
  sense physical stimuli
• When a human co-worker is present at the
  worksite, the teleoperator has the opportunity to
  interact indirectly with the co-worker through
  the robot, which may be considered an extension
  of the teleoperator's own body
• Skill mappings of dexterous robots and human
  workers
• Timeline optimization
• task distribution between crewmembers
  sequence of tasks to be performed

From the co-worker's point of view, interacting
with a teleoperated Robonaut is comparable to
interacting with a human EVA crewmember
11
Data Recorded

• Videotape of team during task performance
• Robot wrist forces/torques
• Socket contact forces/torques
• Task completion times/total lapsed time
• Voice communications between team members

12
Timeline Analysis Results
13
Timeline Analysis Results
14
Contact Forces/Torques
Maximum contact forces and torques can be used to
quantify the risk of hardware damage or failure
due to excessive momentary peak loads Both
control algorithms and human-robot task
distribution can be used to mitigate undesired
forces/torques
15
Conclusions
Primary purpose of this experiment was to
demonstrate the merit of teaming EVA humans with
NASAs latest robotic technologies to perform
orbital assembly of space structures Experiments
tested the limits of both robotics and
teleoperation, demonstrating new extravehicular
robotic (EVR) capabilities and the feasibility of
performing more tasks telerobotically

• Follow-on high-fidelity evaluations should
  include
• more sophisticated gravity compensation
• more realistic mobility
• communication time delays
• realistic lighting conditions