Human-in-the-Loop
Planning and Decision Support
AAAI
2015 Tutorial Forum: Tutorial SP4
Subbarao Kambhampati, Arizona State University
Kartik Talamadupula, IBM Research
Sunday,
01/25/2015, 2:00pm – 6:00pm
Location: Texas 1 Ballroom
Hyatt Regency Austin (Barton Springs)
Endowing an automated agent with
the ability to "plan" -- i.e., converting its high-level goals into
an executable course of action -- has been a long-standing quest in Artificial
Intelligence. For much of the history of
automated planning, the dominant research theme has been efficient synthesis of
plans under increasingly expressive system dynamics (classical, temporal,
stochastic etc.).
An implicit assumption
underlying this research has been that the planner's responsibilities start
with taking a complete specification, and end with giving out a complete course
of action. This assumption is no longer valid when humans are part of the
decision making loop, as is the case in an increasing number of decision
support and human-machine teaming scenarios.
In this tutorial, we will
identify the research challenges in human-in-the-loop planning, including the
need to interpret the goals/intentions of the humans in the loop, the need to
support continual planning and replanning, the need to unobtrusively support
team-decision making, and above all the need to handle pervasive incompleteness
in the domain models as well as problem specification. We will then survey current solutions to
these problems grounded in the context of human-robot teaming, decision support
systems and crowd-sourced planning.
Overview
(Tutorial
Slides (Final version, as given)[PDF] )
1.
Introduction [45min]
(Video)
2.
Challenge:
Interpretation [30min] (Video)
3.
Challenge:
Decision Support
a.
Explicit
Constraints [30min] (Video)
b.
Implicit
Constraints (Preferences) [15+15] (Video for 3b & 3c)
c.
Incomplete
Dynamics
4.
Challenge:
Communication
a.
Excuses & Explanations
b.
Asking for Help
5.
Case Study:
Human-Robot Teaming & Crowdsourced Planning [30min] (Video)
6.
Summary (Video)
Tutorial Slides
(Final version, as given) [PDF]
Subbarao Kambhampati is a professor of
Computer Science at Arizona State University. He is a
AAAI Fellow and AAAI President-Elect, IJCAI Trustee, AAAI 2005 Co-chair, and
IJCAI 2016 Program Chair. He has presented several well-attended and successful
tutorials in past AAAIs.
Email: rao@asu.edu
Kartik Talamadupula is a research staff
member at IBM T.J. Watson Research Center in Yorktown Heights, NY. His
research focuses on the use of automated planners in integrated
human-in-the-loop AI systems. He is a recipient of the SFAz
Fellowship and a University Graduate Fellowship.
Email: krtalamad@us.ibm.com