Bounded Approximate Symbolic Dynamic Programming for Hybrid MDPs.
Luis Gustavo ViannaScott SannerLeliane Nunes de BarrosPublished in: UAI (2013)
Keyphrases
- dynamic programming
- markov decision processes
- state space
- factored mdps
- markov decision problems
- piecewise linear
- reinforcement learning
- optimal policy
- policy evaluation
- dec pomdps
- stereo matching
- finite state
- finite horizon
- greedy algorithm
- markov decision process
- linear programming
- exact and approximate
- multistage
- neural network
- policy search
- decision theoretic planning
- quality guarantees
- average reward
- initial state
- partially observable markov decision processes
- average cost
- optimal control
- coarse to fine
- linear program
- high level