Asymmetric DQN for partially observable reinforcement learning.
Andrea BaiseroBrett DaleyChristopher AmatoPublished in: UAI (2022)
Keyphrases
- partially observable
- reinforcement learning
- state space
- markov decision processes
- partial observability
- dynamical systems
- partially observable domains
- partially observable environments
- decision problems
- hidden state
- partial observations
- function approximation
- markov decision problems
- action models
- reward function
- infinite horizon
- belief space
- reinforcement learning algorithms
- optimal policy
- dynamic programming
- multi agent
- markov decision process
- belief state
- transfer learning
- temporal difference
- domain independent
- continuous state
- partially observable markov decision process
- orders of magnitude
- supply chain
- np hard