Risk-averse control of Markov decision processes with ω-regular objectives.
Rüdiger EhlersSalar MoarrefUfuk TopcuPublished in: CDC (2016)
Keyphrases
- markov decision processes
- risk averse
- state space
- reinforcement learning
- optimal policy
- finite state
- risk neutral
- dynamic programming
- policy iteration
- transition matrices
- risk sensitive
- partially observable
- average cost
- control system
- markov decision process
- decision makers
- average reward
- infinite horizon
- action space
- control strategy
- multiple objectives
- optimal control
- finite horizon
- decision making
- control policies
- function approximation
- utility function
- probability distribution