Combining Markov Decision Processes with Linear Optimal Controllers.
Ekaterina AbramovaAldo FaisalDaniel KuhnPublished in: ICCSW (2011)
Keyphrases
- markov decision processes
- dynamic programming
- reinforcement learning
- average cost
- average reward
- finite horizon
- optimal policy
- action sets
- state space
- total reward
- policy iteration
- finite state
- discounted reward
- partially observable
- decision theoretic planning
- infinite horizon
- reachability analysis
- control policies
- reinforcement learning algorithms
- planning under uncertainty
- action space
- stationary policies
- risk sensitive
- transition matrices
- optimal control
- markov decision process
- factored mdps
- control system
- multistage
- reward function
- model based reinforcement learning
- optimality criterion
- decision processes
- semi markov decision processes
- real time dynamic programming
- interval estimation
- continuous state spaces
- initial state
- expected cost
- optimal strategy
- function approximation
- control strategy
- search algorithm