=============================
REPORT FOR JANUARY-MARCH 2005
=============================

TECHNICAL ACCOMPLISHMENTS

() Extended Optimizer to handle additional features of the "conference
planning" scenario; a list of these new features is posted at
www.cs.cmu.edu/~eugene/Radar/Past/optimize.txt.

() Tested Optimizer on simple "multi-day conference" scenarios, which
gave preliminary evidence of its effectiveness; however, these tests
do not yet provide scalability results, and we plan to run more tests.

() Completed work on the first version of Information Elicitor, which
identifies critical missing pieces of knowledge in an uncertain world
model, and generates related questions about user preferences, room
properties, and distances between rooms.

() Completed work on the first version of Space-Time GUI and
integrated it with Space-Time Module; a list of the main features of
GUI is posted at www.cs.cmu.edu/~eugene/Radar/Past/gui.txt. GUI allows
the user to "collaborate" with Space-Time Module; that is, the user
can make some of the scheduling decisions, and invoke Space-Time
Module to make the other decisions.

PROJECT PLANS

() Extend Optimizer to handle more features of the "conference
planning" scenario; a tentative list of target features is posted at
www.cs.cmu.edu/~eugene/Radar/Short/optimize.txt.

() Continue work on Information Elicitor; a list of target features is
posted at www.cs.cmu.edu/~eugene/Radar/Short/elicit.txt.

() Extend GUI to allow editing the world model, adding and deleting
conference events, and modifying related preferences; a list of target
features is posted at www.cs.cmu.edu/~eugene/Radar/Short/gui.txt.

OBJECTIVES FOR THIS PERIOD

We had the following three objectives for January-March 2005: 

() Produce and test the end-to-end version of Space-Time Module.
() Integrate it with Space-Time GUI and Radar architecture.
() Begin tests involving human subjects.

We have accomplished the first two objectives, but we have not yet
begun work on the third objective. Specifically, we have implemented
and tested the first end-to-end version of Space-Time Module with
Space-Time GUI, and provided a mechanism for adding it to the Radar
architecture. The preliminary experiments have shown that scheduling a
conference using Space-Time Module is much faster than scheduling it
by hand, and that the quality of the resulting schedule is higher than
that of a manually constructed schedule.

OBJECTIVES FOR NEXT PERIOD

() Develop advanced mechanisms for optimization and preference
elicitation. In particular, investigate advanced techniques for
handling uncertainty in the world representation, and incorporate
game-theoretic techniques into Information Elicitor. We expect that
this work will lead to generating better conference schedules based on
incomplete world knowledge, and that it will allow faster and more
accurate elicitation of the missing data.

() Begin tests involving human subjects, in collaboration with Aaron
Steinfeld, who is responsible for the test design. We expect that this
work will provide data on the effectiveness, scalability, and
user-friendliness of the implemented system, and that it will help to
plan work on the next version of Space-Time Module.

MEETINGS AND PRESENTATIONS

We have participated in the IET visits and other internal meetings,
but we have not participated in any external meetings. We made a
presentation during the IET visit on February 17; this presentation is
posted at www.cs.cmu.edu/~eugene/Radar/talks.html. We also gave a demo
of Space-Time Module during the System Development meeting on March 31.