SOS: Safe, Optimal and Small Strategies for Hybrid Markov Decision Processes.
Pranav AshokJan KretínskýKim Guldstrand LarsenAdrien Le CoëntJakob Haahr TaankvistMaximilian WeiningerPublished in: CoRR (2019)
Keyphrases
- markov decision processes
- dynamic programming
- average cost
- action sets
- finite horizon
- average reward
- finite state
- state space
- optimal policy
- policy iteration
- reinforcement learning
- optimal strategy
- reachability analysis
- transition matrices
- discounted reward
- decision theoretic planning
- total reward
- stationary policies
- infinite horizon
- stochastic games
- risk sensitive
- planning under uncertainty
- partially observable
- reinforcement learning algorithms
- action space
- decision processes
- model based reinforcement learning
- markov decision process
- long run
- factored mdps
- state and action spaces
- stochastic shortest path
- control policies
- function approximation
- linear programming
- optimal solution