Belief MCTS

The Belief MCTS solver allows MCTS to be used to solve POMDPs by transforming the POMDP to an MDP on the belief space.

MCTS.BeliefMCTSSolver — Type

BeliefMCTSSolver(mcts_solver, updater)

The belief mcts solver solves POMDPs by modeling them as an MDP on the belief space. The updater is used to update the belief as part of the belief MDP generative model.

Example:

using ParticleFilters
using POMDPModels
using MCTS

pomdp = BabyPOMDP()
updater = SIRParticleFilter(pomdp, 1000)

solver = BeliefMCTSSolver(DPWSolver(), updater)
planner = solve(solver, pomdp)

simulate(HistoryRecorder(max_steps=10), pomdp, planner, updater)

source