نویسندگان
1 دانشجوی کارشناسی ارشد، دانشگاه صنعتی مالک اشتر
2 دانشیار، دانشگاه صنعتی مالک اشتر
3 مربی، دانشگاه صنعتی مالک اشتر
چکیده
کلیدواژهها
عنوان مقاله [English]
نویسندگان [English]
Although existing Planning methods can plan under uncertainty and decentralize situation, most of
them malfunction in some complicated conditions of command and control scenarios such as real time decision
making, need accurate planning, bounded communication between agents, dynamic worlds and partially
observable environments. Among suitable models for these situations, we can consider extended models
of DEC-POMDPs such as MAOP-COMM that can handle these conditions. It is possible to improve
MAOP-COMM model to do planning for agents with double precision. In this paper we have improved the
algorithm of MAOP-COMM model by upgrading value function heuristic and using "two steps lookahead" in
the strategy of finding best policy and making correct decision. Improved algorithm performs online planning
for agents in a multi agent system in uncertain condition with better performance and high percent of
correct decision making. We experiment resulted algorithm on Grid Soccer benchmark. The results obtained
prove efficiently of proposed improvements.
کلیدواژهها [English]
)
