two additional project topics (possible ones)

[Subbarao Kambhampati <rao@asu.edu>]

Here are two additional topics for projects:

Topic 12: Develop and evaluate a multi-valued fluent SAT
encoding based on network flow ideas. Menkes van den Briel is working
with a novel IP encoding of planning problems that look at planning
problems as multiple parallel network flows (one flow per multi-valued
fluent) + interactions between those flows. We would like a student to
replicate his encoding for SAT (Menkes has the encoding for IP). Can
start with Blackbox. 
 
Potential RA contact: Menkes van den Briel
(menkes@asu.edu)
 
Topic 13: Make a regression planner do as well as the state of the art
progression planners.  You should start with something like AltAlt
and make it use one or more of (1) EBL analysis to compute higher order
mutexes (2) forced goal orderings (3) YAHSP style jumps and (4) PDB
heuristics for getting higher order mutexes­to see if regression planners
can be made competitive with best progression planenrs. 
 
Potential RA contacts: Dan Bryce
(dan.bryce@asu.edu) 


Rao