Temporal logic control of general Markov decision processes by approximate policy refinement.
Sofie HaesaertSadegh Esmaeil Zadeh SoudjaniAlessandro AbatePublished in: CoRR (2017)
Keyphrases
- markov decision processes
- temporal logic
- optimal policy
- policy iteration
- policy evaluation
- model checking
- finite state
- factored mdps
- markov decision process
- state and action spaces
- state space
- action space
- finite horizon
- average reward
- decision processes
- average cost
- infinite horizon
- modal logic
- reinforcement learning
- partially observable
- dynamic programming
- reward function
- transition matrices
- control policies
- reinforcement learning algorithms
- decision theoretic planning
- decentralized control
- control policy
- markov decision problems
- planning under uncertainty
- least squares
- linear temporal logic
- temporally extended
- decision problems
- long run
- discounted reward
- temporal difference
- verification method
- control strategy
- stationary policies
- total reward
- action selection
- initial state