This Julia package is a wrapper for my modified APPL code that solves rho-POMDPs.
The code can be called as follows:
sarisa(pomdpx_file::String, timeout::Real=30, variant::String="MN"; filename="temp.policy")`
timeout is expressed in seconds. There are three main
MNMax-norm reward function.
GUGuess reward function.
TH_90Threshold reward function with cutoff at
||b||_inf = 0.9.
There are two more allowed
variant strings, but these use the unimproved lower bound. The only reason they are included is to show the value of using my improved lower bound.
MBMax-norm with bad (unimproved) lower bound.
TB_90Guess reward function with bad (unimproved) lower bound.
The resulting policy file is saved to
11 months ago