Solving average cost Markov decision processes by means of a two-phase time aggregation algorithm.
Edilson Fernandes de ArrudaMarcelo D. FragosoPublished in: Eur. J. Oper. Res. (2015)
Keyphrases
- markov decision processes
- average cost
- dynamic programming
- policy iteration
- finite state
- optimal policy
- model based reinforcement learning
- average reward
- reinforcement learning
- transition matrices
- optimal solution
- computational complexity
- linear programming
- np hard
- control policy
- infinite horizon
- semi markov decision processes
- state space
- learning algorithm
- planning under uncertainty
- decision processes
- stochastic shortest path
- objective function
- finite horizon
- partially observable
- long run
- approximate dynamic programming
- decision theoretic planning
- convergence rate
- linear program
- data mining