Fast Approximate Dynamic Programming for Infinite-Horizon Continuous-State Markov Decision Processes.
Mohamad Amin Sharifi KolarijaniG. F. MaxPeyman Mohajerin EsfahaniPublished in: CoRR (2021)
Keyphrases
- approximate dynamic programming
- continuous state
- markov decision processes
- infinite horizon
- reinforcement learning
- average cost
- policy iteration
- finite state
- action space
- optimal policy
- dynamic programming
- partially observable markov decision processes
- finite horizon
- control policies
- state space
- state dependent
- partially observable
- dec pomdps
- optimal control
- robot navigation
- control policy
- function approximation
- markov decision process
- average reward
- planning under uncertainty
- reinforcement learning algorithms
- machine learning
- model free
- decision problems
- linear program
- learning algorithm
- initial state
- stochastic games
- state action
- multi agent
- reward function
- step size
- planning problems
- policy gradient