Solving Risk-Sensitive POMDPs With and Without Cost Observations.
Ping HouWilliam YeohPradeep VarakanthamPublished in: AAAI (2016)
Keyphrases
- risk sensitive
- markov decision problems
- markov decision processes
- average cost
- partially observable
- linear programming
- reinforcement learning
- optimal policy
- state space
- optimal control
- utility function
- decision theoretic
- markov decision chains
- dynamic programming
- decision processes
- policy iteration
- expected utility
- state transitions
- finite horizon
- finite state
- long run
- transition probabilities
- control policies
- decision problems
- total cost
- infinite horizon
- model free
- partially observable markov decision processes
- queueing networks
- steady state
- linear program
- multistage
- reinforcement learning algorithms
- evolutionary algorithm
- learning algorithm
- initial state
- mobile robot
- planning problems
- function approximators
- probability distribution
- average reward
- dynamical systems
- optimal solution
- multi agent