Bounded Approximate Symbolic Dynamic Programming for Hybrid MDPs.
Luis Gustavo ViannaScott SannerLeliane Nunes de BarrosPublished in: CoRR (2013)
Keyphrases
- dynamic programming
- markov decision processes
- factored mdps
- state space
- reinforcement learning
- piecewise linear
- optimal policy
- approximate dynamic programming
- markov decision problems
- dec pomdps
- infinite horizon
- optimal control
- policy evaluation
- dynamic programming algorithms
- exact and approximate
- finite state
- symbolic representation
- hybrid learning
- multistage
- policy iteration
- markov decision process
- average cost
- stereo matching
- linear programming
- finite horizon
- partially observable
- probabilistic planning
- coarse to fine
- symbolic data
- single machine
- action space
- average reward
- long run
- decision diagrams
- high level