Hunter Hale

Previous Week:

Nailed down design for FLAIRS paper motivation and experiment.

Constructed Several Block World components.

Implemented C++ version of dynamic rebuilding of decomps (in progress eta today)

Expanding visualizer to show agent movement and discovery of block world components, also enhanced visualizer with better exception handling.

Finished some typed proofs for Dr. Barnes – still have a few chapters to type up.

Planned Work

Complete and Run FLAIRs paper experiment target Wednesday

Finalize dissertation committee and submit first set of paperwork – target Wednesday

Filled in Dissertation outline – target required for finalization of committee members.

More proofs for Dr. B

Draft of FLAIRS paper target Friday

Interface work on DEACCON – target as time permits

ToDo:
DEACCON work – interface more granular loading of models features for FLAIRS papers
Committee and PhD proposal setup targeting week of December 7th

FLAIRS paper
Hyp: It is possible to discover a navigation mesh using information gained from one or more agents moving through the world.

Applications: Simulations of Search and rescue in unknown environments, more “fair” agents for games and simulations, others

Approach: Crib some of the D* working with unknown areas and combine that with dynamic rebuilding of decomps. Assume world is empty then rebuild when this assumption is invalidated.

Validation and Experimentation: Compare the error levels present in a agents generated decomposition (incorrectly classified space, missed gateways, average path lengths) to the “optimum decomposition” in a time lapsed manor. How quickly does agent discovery converge on optimum decomp?