Allocation of Continuous Resources in Agent Systems

This monograph concentrates on reasoning techniques for
intelligent, autonomous agent systems. In particular,
it focuses on
planning techniques for both single and multi-agent
systems acting
in uncertain domains. In modeling these domains, two
types of
uncertainty are considered: (i) the outcomes of agent
actions are
uncertain and (ii) the amount of resources consumed
by agent
actions is uncertain and only characterized by
continuous
probability density functions. For solving planning
problems
modeled in these domains the monograph proposes a
class of
efficient algorithms that provide guarantees on the
solution quality.
The experimental evaluation of the proposed
algorithms shows up
to three orders of magnitude speedups in solving
single agent
planning problems and up to one order of magnitude
speedup in
solving multi-agent planning problems. Additionally,
it is
demonstrated how the proposed algorithms allow for
more efficient
rescue operation in a large-scale disaster simulation.

Autorentext

Janusz Marecki is a scientist at IBM T.J. Watson Center and avisiting professor at the Academy of Computer Science (WSIZ). He holds a PhD fromUniversity of Southern California and a DrSc from Institute of InformationInfrastructure in Lviv. Janusz wrote over 50 publications and 2 patents and conducts research indecision & game theory.

Klappentext

This monograph concentrates on reasoning techniques for intelligent, autonomous agent systems. In particular,it focuses on planning techniques for both single and multi-agentsystems acting in uncertain domains. In modeling these domains, twotypes of uncertainty are considered: (i) the outcomes of agentactions are uncertain and (ii) the amount of resources consumedby agent actions is uncertain and only characterized bycontinuous probability density functions. For solving planningproblems modeled in these domains the monograph proposes aclass of efficient algorithms that provide guarantees on thesolution quality. The experimental evaluation of the proposedalgorithms shows up to three orders of magnitude speedups in solvingsingle agent planning problems and up to one order of magnitudespeedup in solving multi-agent planning problems. Additionally,it is demonstrated how the proposed algorithms allow formore efficient rescue operation in a large-scale disaster simulation.