F R O N T M A T T E R
Overcoming Remote Perception Challenges to Sup-
port Decision Making in Human-Robot Teams
Martin Voshell
The Ohio State University
voshell.2@osu.edu
Dr. Flip Phillips
Skidmore College
flip@skidmore.edu
Dr. David Woods
The Ohio State University
woods.2@osu.edu
Our current research extends upon work we first we presented at NDM6 envision-
ing human robot coordination in future operations. We discussed the expanding roles that
tele-operated and semi-autonomous robots played in exploring remote environments and
the many new coordination challenges and cognitive demands that accompany such
technological change when introduced into existing workplaces. Through our experi-
ences, work analysis, and naturalistic observations working with robotic search and rescue
groups and disaster response teams we have seen that many basic perceptual and coordina-
tion links between the people and the remote environments are broken.
There are many challenges in designing successful human-robot systems and
integrating teams and multiple problem holders, the most fundamental coordination
challenge is overcoming the significant decoupling between the natural linkage between
the perception and action of robot handlers and the information coming to them from the
remote robot platforms. Robot handlers specifically experience many difficulties: poor
lighting robs color, can give false specularity cues, and greatly hinders texture discrimina-
tion; there is no coordination between the heading and gaze of the robot; with no frame of
reference for body awareness there is profound misperception of depth, speed, and scale
of obstacles and passages; all of which significantly impact work capability.
Our ecologically-minded goal became to simply investigate how we could better
support the perception of affordances in remote environments by remote operators. We
created a high fidelity virtual environment to use for scaled world simulations and to test
new display concepts inspired by our field observations and interactions in attempt to
reduce these breakdowns and gain a better understanding of those very ambiguities. We
can quickly prototype interface designs and task situations in the virtual environment and
have begun to evaluate them using a novel and robust, high-level quantitative analysis
framework coupled with process tracing and observational methodologies. In the current
study we are focusing on improving the perception-action coupling between the robot
handler and the mobile robotic systems by capturing and preserving multiple camera
arrangements and multiple problem holders’ points of view to aid navigation and
coordination.
20 January, 2005
B O D Y
B A C K M A T T E R
submitted to NDM7 (http://www.ndm7.org/)