Switching Q-learning in partially observable Markovian environments.
Hiroyuki KamayaHaeyeon LeeKenichi AbePublished in: IROS (2000)
Keyphrases
- partially observable
- reinforcement learning
- state space
- reward function
- markov decision processes
- reinforcement learning algorithms
- decision problems
- optimal policy
- partially observable environments
- partial observability
- dynamical systems
- infinite horizon
- partial observations
- markov decision problems
- function approximation
- action models
- learning algorithm
- belief state
- partially observable domains
- heuristic search
- markov chain
- partially observable markov decision process
- orders of magnitude
- decision making
- action space
- partially observable markov decision processes
- model free
- planning domains
- state variables
- planning problems
- multi agent
- markov decision process
- policy iteration
- learning agent
- long run
- dynamic environments
- domain specific