=============================
REPORT FOR JANUARY-MARCH 2006
=============================

TECHNICAL ACCOMPLISHMENTS

() Made several algorithmic improvements to the scheduler, which led
to significant increase in both speed and schedule quality. Also, ran
extensive experiments on the performance of the scheduler under
various conditions, and identified the values of control parameters
that lead to near-optimal performance.

() Extended a mechanism for generating "common-sense" assumptions
about user preferences and available resources, and integrated it with
the rest of the Space-Time module.

() Integrated the Space-Time module with CMRadar, which has been a
joint work with the CMRadar group. The resulting integrated system
supports the analysis of trade-offs between the cost of getting new
rooms and the resulting schedule improvements, and makes a
near-optimal selection of additional rooms.

() Began work on a mechanism for learning of elicitation strategies;
in particular, developed a representation for the learned knowledge.
Also, began work on a mechanism for learning "common-sense"
assumptions based on the available information.


IMPROVEMENTS TO PROTOTYPE

() Revamped the representation of conference-scheduling scenarios, and
developed an initial version of the API for accessing the world model
from other parts of the Radar architecture.

() Improved the integration of the Space-Time module with the
Space-Time GUI; in particular, provided time estimates for all slow
operations, which allow the display of progress bars, and also enabled
the user to interrupt these slow operations.

() Extended the procedure for checking the syntactic and semantic
correctness of conference-planning scenarios.


OBJECTIVES FOR NEXT PERIOD

() Extend the optimization algorithm to handle room costs, movable
equipment, services, and vendors.

() Continue work on the mechanisms for learning of elicitation
strategies and "common-sense" assumptions.

() Implement an initial version of the "contingency scheduling"
system, which will allow construction of alternative schedules for
multiple world models. Also, begin work on a mechanism for learning of
contingency-scheduling strategies.


CONCERNS

() The main concern is the same as in the previous quarterly report;
specifically, we need to develop GUI tools for the editing of
conference-planning scenarios, which may require collaboration with
GUI researchers.  Currently, the construction of such scenarios
requires manual editing of the underlying XML representation, which is
a slow process.