id	author	title	date	pages	extension	mime	words	sentences	flesch	summary	cache	txt
work_7wyyn4jlmzflvnev4k7qonrx2i	Christopher Amato	Modeling and Planning with Macro-Actions in Decentralized POMDPs	2019	43	.pdf	application/pdf	20076	1789	67	Decentralized partially observable Markov decision processes (Dec-POMDPs) are general models for decentralized multi-agent decision making under uncertainty. solutions to be generated for significantly longer horizons and larger state-spaces than previous Dec-POMDP methods. Singh, 1999), have explored using higher-level, temporally extended macro-actions (or options ) to represent and solve problems, leading to significant performance improvements in multi-agent case by introducing a Dec-POMDP formulation with macro-actions modeled extended to multi-robot systems: our methods naturally bridge Dec-POMDPs and multirobot coordination, allowing principled decentralized methods to be applied to real domains. To solidify this bridge, we describe a process for creating a multi-robot macro-action DecPOMDP (MacDec-POMDP) model, solving it, and using the solution to produce a set of Dec-POMDPs (Bernstein et al., 2002) generalize POMDPs1 (Kaelbling, Littman, & Cassandra, 1998) and MDPs2 (Puterman, 1994) to the multi-agent, decentralized setting. is, given a current belief state, b, and a policy of option-based macro-actions, µ, the value	./cache/work_7wyyn4jlmzflvnev4k7qonrx2i.pdf	./txt/work_7wyyn4jlmzflvnev4k7qonrx2i.txt