Pat Callaghan
Hi! I’m Pat, and I’m a PhD student in the
Robotics Institute
at Carnegie Mellon University where I have the good fortune of being co-advised
by Professors Henny Admoni and
Reid Simmons.
Prior to becoming a PhD student, I was a Master's student in the RI. During
that time, I was co-advised by Professors Henny Admoni and
Oliver Kroemer.
Before enrolling at CMU, I worked on the navigation and planning software for
MoonRanger under Professors
Red Whittaker and David Wettergreen, and I also worked on the
Iris (originally
CubeRover) project's Systems
Engineering team. Before robotics, I taught 9th grade Humanities and coached
boys' lacrosse and ice hockey at Culver Academies. I earned a dual degree in
English and Economics from the University of Virginia.
Email  / 
CV  / 
Google Scholar
|
|
Research
These days, I'm especially interested in figuring out how robot learners can help
make people more
efficacious
teachers of skills and personal preferences. In particular, I am working on ways to:
- model misunderstandings that learners and teachers have of each other, and
- incorporate such models into the robot's reasoning and feedback generation
processes.
Ideally, my work will reduce the cognitive demands currently placed upon human
teachers without sacrificing the interpretability of the robot's internal
models.
|
|
Exploring Diverse Interaction Types for Human-in-the-Loop
Robot Learning.
Patrick Callaghan.
Carnegie Mellon University, 2023
Document
My Master's of Science in Robotics (MSR) work
investigated different interaction types and reward functions
for representing human preferences.
|
|
INQUIRE: INteractive Querying for User-aware Informative
REasoning.
Tesca Fitzgerald,
Pallavi Koppol,
Patrick Callaghan,
Russell Q. Wong,
Reid Simmons,
Oliver Kroemer,
Henny Admoni.
Conference on Robot Learning, 2022
Github  / 
Paper
We studied how an interactive agent might actively select from multiple
forms of human-provided feedback to learn a human's preferences.
|
|
Image and Lidar Dataset of the West Desert Sinkhole: An Analog for
Steep-walled Planetary Pits.
Jordan Ford,
Patrick Callaghan,
Uland Wong,
Heather Jones,
Chuck Whittaker,
William "Red" Whittaker.
International Planetary Caves Conference, 2020
Paper
We went to the desert and collected data to model the moon's massive pits (which might one
day serve as outposts for astronauts).
|
Thanks to Jon Barron for making his website's
source code
available to the rest of us!
|
|