Implements anytime error minimization search (AEMS). Specifically, this is AEMS2, which generally outperforms AEMS1.

Example with some optional arguments:

AEMSSolver(max_iterations = 100, action_selector = :L)

Fields:

• max_iterations Maximum node expansions per action.

• max_time Maximum time (in seconds) to spend per action.

• updater Updater used to propagate belief nodes. Defaults to DiscreteUpdater.

• lower_bound Subtype of Policy. Defaults to fixed action policy.

• upper_bound Subtype of Policy. Defaults to FIB policy.

• root_manager Allowable values are :clear, :belief, or :user. Defaults to :clear.

• action_selector Allowable values are :U or :L. Defaults to :U

Clears existing action/observation graph.

Returns tuple (nb, na), numbers of belief and action nodes in graph