POPGym: Benchmarking Partially Observable Reinforcement Learning.
Steven D. MoradRyan KortvelesyMatteo BettiniStephan LiwickiAmanda ProrokPublished in: ICLR (2023)
Keyphrases
- partially observable
- reinforcement learning
- state space
- markov decision processes
- partial observability
- partially observable domains
- partial observations
- dynamical systems
- decision problems
- hidden state
- partially observable environments
- function approximation
- markov decision problems
- belief space
- reward function
- action models
- dynamic programming
- infinite horizon
- reinforcement learning algorithms
- learning algorithm
- multi agent
- temporal difference
- optimal policy
- partially observable markov decision process
- fully observable
- learning capabilities
- special case
- action space
- action selection
- model free
- optimal control
- hidden markov models
- transfer learning
- np hard