From: Alexander Yates (ayates_at_cs.washington.edu)
Date: Tue May 27 2003 - 10:26:34 PDT
Review of "Temporal Planning with Mutual Exclusion Reasoning," by Smith and
Weld.
This paper describes how to use mutex relations when planning with actions
that can have arbitrary durations.
The authors add two key components to the mutex reasoning of TGP, their
planner. First is the ability to make actions and propositions mutex so
that if a proposition could become true in the middle of some action's
execution and violate that action's preconditions or effects, the action and
proposition can be labeled mutex. The other new component are the cmutexes
and their labels. Although they need to be approximated for efficiency
reasons, these conditions, in conjunction with action/prop. mutexes, provide
flexibility to express more mutexes than would otherwise be possible.
I'm not quite sure what figure 5 shows.
It doesn't seem like TGP would have the right level of expressivity for NASA
rovers (assuming that's what they're aiming for, could be wrong), since
these things need to think about resource consumption (continuous change)
and actions with variable duration and such. The paper talks about
extending it in various ways, but it's not clear to me that this framework
can handle these kinds of things.
One possible algorithm extension for speeding things up is using EBL (which
could get kinda complicated with the different time intervals) & DDB, which
seemed pretty helpful for STAN.
Alex
This archive was generated by hypermail 2.1.6 : Tue May 27 2003 - 10:26:48 PDT