Parallel asynchronous distributed computations of optimal control in large state space Markov Decision processes.
Bruno ScherrerPublished in: ESANN (2003)
Keyphrases
- markov decision processes
- optimal control
- state space
- dynamic programming
- reinforcement learning
- risk sensitive
- infinite horizon
- optimal policy
- policy iteration
- average cost
- control problems
- finite state
- reinforcement learning algorithms
- transition matrices
- finite horizon
- markov chain
- policy iteration algorithm
- partially observable
- decision theoretic planning
- multi agent
- heuristic search
- function approximation
- reachability analysis
- reward function
- control strategy
- action sets
- particle filter
- markov decision process
- state variables
- continuous state spaces
- markov decision problems
- action space
- average reward
- macro actions
- initial state
- sufficient conditions
- multistage
- linear program
- state abstraction
- search space
- state and action spaces
- decision problems
- production planning
- model free
- data mining