Automatic Construction of Temporally Extended Actions for MDPs Using Bisimulation Metrics.
Pablo Samuel CastroDoina PrecupPublished in: EWRL (2011)
Keyphrases
- automatic construction
- temporally extended
- decision theoretic planning
- markov decision processes
- planning under uncertainty
- reinforcement learning
- qualitative simulation
- markov decision problems
- finite state
- dec pomdps
- state space
- factored mdps
- probabilistic planning
- temporal aspects
- natural language
- state and action spaces
- decision processes
- decision theoretic
- linear programming
- average reward
- action space
- temporal logic