Reachability and Safety Objectives in Markov Decision Processes on Long but Finite Horizons.
Galit Ashkenazi-GolanJános FleschArkadi PredtetchinskiEilon SolanPublished in: J. Optim. Theory Appl. (2020)
Keyphrases
- markov decision processes
- state space
- state and action spaces
- infinite horizon
- finite state
- reinforcement learning
- optimal policy
- policy iteration
- reachability analysis
- finite horizon
- decision theoretic planning
- partially observable
- dynamic programming
- transition matrices
- average reward
- heuristic search
- reinforcement learning algorithms
- stationary policies
- action space
- decision processes
- average cost
- semi markov decision processes
- action sets
- risk sensitive
- multiple objectives
- markov decision process
- finite number
- state abstraction
- reward function
- model based reinforcement learning
- markov decision problems
- planning under uncertainty
- machine learning
- real valued
- interval estimation