From: Parag (parag_at_cs.washington.edu)
Date: Fri May 30 2003 - 10:44:31 PDT
Sapa: A Domain-Independent Heuristic Metric
Temporal Planner
- Minh B. Do & Subbarao Kambhampati
The paper presents a domain-independent forward
chaining planner which works in metric temporal
domains. It derives its heuristics
from the relaxed planning graph representation.
The main ideas in the paper are:
Formulation of a forward chain planner which would handle
temporal and metric constraints. The authors
propose an action/goal/state representation
for these kind of domains, which could be
used to search for the goal state efficiently.
Use of the relaxed planning graph representation
to derive the heuristics for forward chaining
- authors talk about a handful of heuristics -
some of them admissible and others which are
non-admissible. The idea of ignoring metric
constraints in the initial calculation of
the heuristic values and then incorporating
them at a later stage seemed quite interesting.
The experimental section seemed rich enough.
The authors allude to the fact that there are
other richer representations for temporal domains
than their's. A slightly more detailed analysis of
where the difference lies and if the techniques
could be extended to such domains would have helped.
There are many directions for the future work.
First would be to explore if one could extend
the current framework to richer domains as
represented by zeno. One could also delve into
why certain heuristics perform better than
others(as described in the experimental section)
and if there is any relation between the heuristic
being used the specific domain one is looking at.
This archive was generated by hypermail 2.1.6 : Fri May 30 2003 - 10:44:31 PDT