Q-learning for Markov decision processes with a satisfiability criterion.
Suhail M. ShahVivek S. BorkarPublished in: Syst. Control. Lett. (2018)
Keyphrases
- markov decision processes
- optimal policy
- reinforcement learning
- reinforcement learning algorithms
- state space
- policy iteration
- stochastic shortest path
- markov games
- discounted reward
- reward function
- dynamic programming
- finite state
- continuous state spaces
- finite horizon
- transition matrices
- discount factor
- multi agent
- infinite horizon
- planning under uncertainty
- factored mdps
- reachability analysis
- partially observable
- average reward
- function approximation
- decision theoretic planning
- markov decision process
- markov chain
- action space
- decision processes
- action sets
- state action
- state and action spaces
- learning rate
- state abstraction
- optimal control
- decision problems
- optimality criterion
- markov decision problems
- average cost
- computational complexity
- model free
- real valued
- learning algorithm
- policy evaluation
- initial state
- cooperative
- temporal difference
- action selection