Safety-Constrained Reinforcement Learning for MDPs.
Sebastian JungesNils JansenChristian DehnertUfuk TopcuJoost-Pieter KatoenPublished in: TACAS (2016)
Keyphrases
- reinforcement learning
- markov decision processes
- state space
- optimal policy
- reinforcement learning algorithms
- control problems
- partially observable
- model free
- function approximation
- multi agent
- markov decision process
- temporal difference
- state and action spaces
- continuous state and action spaces
- machine learning
- action space
- policy iteration
- finite state
- policy search
- learning algorithm
- policy evaluation
- factored markov decision processes
- model based reinforcement learning
- markov games
- dynamic programming
- supervised learning
- approximate dynamic programming
- markov decision problems
- temporal difference learning
- optimal control
- infinite horizon
- planning under uncertainty
- reinforcement learning methods
- reward function
- state abstraction
- average reward
- reinforcement learning problems
- dynamical systems
- action sets
- finite horizon
- learning process