Solving Markov Decision Processes with Partial State Abstractions.
Samer B. NashedJustin SvegliatoMatteo BrucatoConnor BasichRod GrupenShlomo ZilbersteinPublished in: ICRA (2021)
Keyphrases
- markov decision processes
- state space
- transition matrices
- semi markov decision processes
- partially observable
- optimal policy
- action space
- dynamic programming
- state abstraction
- markov decision problems
- policy iteration
- markov decision process
- reinforcement learning
- state variables
- real time dynamic programming
- finite state
- finite horizon
- planning under uncertainty
- reachability analysis
- reinforcement learning algorithms
- decision theoretic planning
- risk sensitive
- infinite horizon
- factored mdps
- markov chain
- discounted reward
- heuristic search
- state and action spaces
- total reward
- average cost
- machine learning
- learning algorithm
- action sets
- reward function
- partially observable markov decision processes