Light robustness in the optimization of Markov decision processes with uncertain parameters.
Peter BuchholzDimitri ScheftelowitschPublished in: Comput. Oper. Res. (2019)
Keyphrases
- markov decision processes
- state space
- finite state
- optimal policy
- reinforcement learning
- transition matrices
- dynamic programming
- reinforcement learning algorithms
- action space
- planning under uncertainty
- reachability analysis
- policy iteration
- decision theoretic planning
- finite horizon
- decision processes
- decision making
- model based reinforcement learning
- factored mdps
- average cost
- partially observable
- average reward
- sensitivity analysis
- state abstraction
- infinite horizon
- markov decision process
- action sets
- discounted reward
- state variables