This package defines a policy for rolling out a pre-defined sequence of actions to an MDP or POMDP defined with the POMDPs.jl package.
A PlaybackPolicy can be constructed in two ways:
PlaybackPolicy(actions, backup_policy::Policy)- whereactionsis a sequence of actions to output andbackup_policyis the policy used if the MDP has not terminated by the time all the prescribed actions have been used.PlaybackPolicy(actions, logpdfs::Vector{Float64}, backup_policy::Policy)- The only difference for this constructor is the ability to specify the log-probability of the prescribed actions aslogpdfs. Use this in combination with algorithms that require the probability of actions.
Once defined, get an action using POMDPs.action(p::PlaybackPolicy, s)
Maintained by Anthony Corso (acorso@stanford.edu)