Information Technology Reference
In-Depth Information
w. r. t. HMAP. Therefore, it is not a big disadvantage of using HMAP instead of PLM
for problems which have a single task in the initial plan. Not surprisingly, the perfor-
mance improvements will increase dramatically with the number of tasks in the initial
plan. Our experiments proved that when there is a causal interaction between tasks in
the plan, the Ind decomposition technique is more efficient than the Dep decomposi-
tion technique such as in the UM-Translog domain, where the Ind technique achieves
an average improvement of
w. r. t. the Dep technique as documented in table 1.
Although, satellite domain does not benefit significantly from the landmark prepro-
cessing technique due to the shallow decomposition hierarchy, it achieves good perfor-
mance with decomposition techniques (either Dep or Ind technique) as depicted in
table 2.
22%
6Con lu ion
We have presented a new hybrid approach that integrates the hierarchical landmark
preprocessing technique with MAP. Our approach enables us to break up the plan-
ning problem into a set of clusters using two different techniques; Dependent and
Independent . It guarantees that: (1) the set of agents work independently, (2) the in-
dividually constructed plans are merged successfully in order to generate a global plan
without additional refinement in any individual plan, and (3) the problems are solved in
shorter time. We have performed a number of experiments on our representation frame-
work on exemplary problem specifications for two hierarchical domains in which the
HMAP approach competed with a planner “with and without” preprocessing. These
results give evidence for the practical relevance of our approach.
Acknowledgements. This work is done within the Transregional Collaborative Re-
search Centre SFB/TRR 62 “Companion-Technology for Cognitive Technical Systems”
funded by the German Research Foundation (DFG).
References
1. Biundo, S., Schattenberg, B.: From abstract crisis to concrete relief (a preliminary report on
combining state abstraction and HTN planning). In: Proc. of ECP, pp. 157-168 (2001)
2. Bradley, J., Edmund, H.: Theory for coordinating concurrent hierarchical planning agents
using summary information. In: Proc. of AAAI, pp. 495-502 (1999)
3. Corkill, D.: Hierarchical planning in a distributed environment. In: Proc. of IJCAI, pp. 168-
175 (1979)
4. desJardins, M., Wolverton, M.: Coordinating a distributed planning system. Journal of AI
Magazine 20(4), 4553 (1999)
5. Elkawkagy, M., Schattenberg, B., Biundo, S.: Landmarks in hierarchical planning. In: Proc.
of ECAI, pp. 229-234 (2010)
6. Elkawkagy, M., Bercher, P., Schattenberg, B., Biundo, S.: Exploiting landmarks for hybrid
planning. In: 25th PuK Workshop Planen, Scheduling und Konfigurieren, Entwerfen (2010)
7. Erol, K., Hendler, J., Nau, D.: UMCP: A sound and complete procedure for hierarchical
task-network planning. In: Proc. of AIPS, pp. 249-254 (1994)
Search WWH ::




Custom Search