Decomposition Methods for Solving Finite-Horizon Large MDPs

Algorithm 1

Ameliorated backward induction.

(1)	ABI (In MDP: Out
(2)
(3)	Take
(4)	Repeat
(5)
(6)	For each Do
(7)	//The Deterministic Decision Rule
(8)
(9)	For each Do
(10)	is an optimal policy and is the optimal value vector.
11	Return