A controllability condition for the existence of average optimal stationary policies of linear stochastic systems.
Alessandro N. VargasJoão B. R. do ValPublished in: ECC (2009)
Keyphrases
- stationary policies
- sufficient conditions
- markov decision processes
- stochastic systems
- average cost
- state space
- finite state
- dynamic programming
- markov decision process
- linear program
- lot sizing
- optimal policy
- long run
- infinite horizon
- finite number
- inventory level
- reward function
- fixed point
- markov decision problems
- partially observable
- production planning
- confidence intervals
- markov chain
- initial state
- policy iteration
- total cost
- optimal control
- linear programming
- reinforcement learning