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