Continuous Real Time Dynamic Programming for Discrete and Continuous State MDPs.
Luis Gustavo Rocha ViannaScott SannerLeliane Nunes de BarrosPublished in: BRACIS (2014)
Keyphrases
- continuous state
- real time dynamic programming
- markov decision processes
- action space
- continuous action
- continuous state spaces
- state space
- finite state
- continuous state and action spaces
- reinforcement learning
- policy search
- markov decision problems
- optimal policy
- partially observable markov decision processes
- control policies
- robot navigation
- policy iteration
- dynamic programming
- partially observable
- average reward
- decision processes
- state action
- markov decision process
- reinforcement learning algorithms
- markov chain
- finite horizon
- function approximators
- learning algorithm
- heuristic search
- real valued
- reward function
- action selection
- infinite horizon
- dynamical systems
- planning problems
- state dependent
- average cost
- search space