Policy-Conditioned Uncertainty Sets for Robust Markov Decision Processes.
Andrea TirinzoniMarek PetrikXiangli ChenBrian D. ZiebartPublished in: NeurIPS (2018)
Keyphrases
- markov decision processes
- optimal policy
- policy iteration
- markov decision process
- finite horizon
- infinite horizon
- average cost
- average reward
- reinforcement learning
- partially observable
- reward function
- finite state
- decision processes
- state and action spaces
- action space
- state space
- transition matrices
- dynamic programming
- policy evaluation
- reinforcement learning algorithms
- discounted reward
- model based reinforcement learning
- markov decision problems
- decision theoretic planning
- decision problems
- semi markov decision processes
- expected reward
- robust optimization
- reachability analysis
- policy iteration algorithm
- multistage
- total reward
- control policies
- planning under uncertainty
- factored mdps
- stationary policies
- long run
- partially observable markov decision processes
- state abstraction
- action sets
- belief functions
- initial state