A toy pomdp model for Carbon Storage
Sample usage in REPL:
- Activate with
] activate .
- Run a POMCPOW solver
include("scripts/solve.jl")
The POMDP formulation tracks suitability of various grid locations. Suitability is an average of individual scores for each feature. If you have a feature that you think would help determine suitability, you can add it with the following steps:
- In
src/config.jl
add the name of your feature toFEATURE_NAMES
- In
src/config.jl
add a prior belief of your variable's value:featureName => (mean, variance)
to thePRIOR_BELIEF
dictionary - In
src/config.jl
add action uncertainties to thea_u
dictionary(:action_type, :featureName) => variance
- This means that an action of
:action_type
can inform us of the value of:featureName
withvariance
- This means that an action of
- Include logic on how your feature is to be scored in the
score_component
function, which returns a value 0 - 5