CCSPOMDPs

A toy pomdp model for Carbon Storage

Sample usage in REPL:

1. Activate with ] activate .
2. Run a POMCPOW solver include("scripts/solve.jl")

Adding Features

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:

1. In src/config.jl add a prior belief of your variable's value :featureName => (mean, variance) to the PRIOR_BELIEF dictionary
2. In src/config.jl add action uncertainties to the a_u dictionary (:action_type, :featureName) => variance
   • This means that an action of :action_type can inform us of the value of :featureName with variance
3. Modify the initialize_earth() and CCSPOMDP() functions to include your feature
   • initialize_earth() will require you to write some random initialization for the value of the feature
   • Modify the constructor by just adding the feature name to the feature_names vector
4. Include logic on how your feature is to be scored in the score_component function, which returns a value 1 - 5

Planning Links:

• Excalidraw (Diagrams/Notes)
• SharePoint
• Planning + Action Items
• Progress Report