Generalized Optimality Guarantees for Solving Continuous Observation POMDPs through Particle Belief MDP Approximation.
Michael H. LimTyler J. BeckerMykel J. KochenderferClaire J. TomlinZachary N. SunbergPublished in: CoRR (2022)
Keyphrases
- markov decision problems
- policy iteration algorithm
- markov decision processes
- quality guarantees
- finite state
- belief state
- reinforcement learning
- partially observable
- partially observable markov decision processes
- average cost
- state space
- sequential decision making problems
- average reward
- action space
- approximation methods
- policy iteration
- expected utility
- optimal policy
- linear programming
- point based value iteration
- queueing networks
- dynamic programming
- belief space
- decision theoretic planning
- continuous functions
- markov chain
- stationary policies
- eigenvalue problems
- utility function
- markov decision process
- initial state
- infinite horizon
- planning under uncertainty
- continuous state spaces
- dynamical systems
- factored mdps
- continuous state
- optimal solution
- dynamic programming algorithms
- piecewise constant
- model checking
- decision theoretic
- belief revision
- linear program
- reward function
- model free
- transition probabilities
- long run
- distributed constraint optimization
- continuous action
- belief functions
- decision problems