Finite Linear Programming Approximations of Constrained Discounted Markov Decision Processes.
François DufourTomás Prieto-RumeauPublished in: SIAM J. Control. Optim. (2013)
Keyphrases
- markov decision processes
- linear programming
- dynamic programming
- state and action spaces
- optimal policy
- linear program
- policy iteration
- stationary policies
- average reward
- average cost
- state space
- finite state
- policy evaluation
- finite horizon
- reinforcement learning
- transition matrices
- infinite horizon
- model based reinforcement learning
- action space
- reachability analysis
- factored mdps
- markov decision problems
- decision processes
- planning under uncertainty
- markov games
- decision theoretic planning
- reinforcement learning algorithms
- partially observable
- optimal solution
- action sets
- np hard
- finite number
- approximation methods
- risk sensitive
- markov decision process
- reward function
- discounted reward
- objective function
- optimal control
- efficient computation