CALP

Contraindicated offline POMDP solver.
Contrained belief state MDP Linear Programming belief set generation Approximate POMDP with Contrainst CPOMDPs are Hard Can’t do DP with pruning: optimal policies may be stochastic Minimax quadratically contained program: computational intractable Contained PBVI struggles with contraint satisfaction CALP Core Idea Recast CPOMDP as a contrained belief-state MDP.
We replace our state-space with our belief space:
S = B s_0 = b_0 You essentially assume here that there is some finite belief space.