From: Christophe Bisciglia (chrisrb_at_cs.washington.edu)
Date: Fri May 23 2003 - 10:32:52 PDT
J.S. Penberthy and D. Weld Temporal Planning with Continuous Change
This paper describes ZENO, a partial order planner that supports temporal
constraints and continuous change.
One of the big ideas in this paper was the expressive language that ZENO
uses to represent problems. Unlike any language I have seen to date,
ZENO.s language allows continuous variables to be quantified (most
importantly time), and constraints to be placed on these variables.
Another key idea was the combination of regression search with linear
programming. At each state, the planner solves a linear program to ensure
consistency. This allows for constraints to be defined as piecewise linear
functions and general inequalities (<=, <, > c, etc) . this is quite
powerful compared to ZENO.s predecessors.
It would have been nice to see more data on ZENO.s performance . the fact
that it solves two problems quickly, and one orders of magnitude slower
(but still reasonably fast) leaves me wondering how it works on both
larger problems, and those in the middle.
I would be curious to see how linear programming optimizations could
improve the planner.s performance.
This archive was generated by hypermail 2.1.6 : Fri May 23 2003 - 10:32:52 PDT