Bingham Policy Parameterization for 3D Rotations in Reinforcement Learning.
Stephen JamesPieter AbbeelPublished in: CoRR (2022)
Keyphrases
- reinforcement learning
- optimal policy
- policy search
- action selection
- markov decision process
- policy evaluation
- partially observable environments
- partially observable
- actor critic
- reinforcement learning problems
- function approximators
- reinforcement learning algorithms
- function approximation
- state space
- markov decision processes
- policy gradient
- state and action spaces
- markov decision problems
- temporal difference
- reward function
- action space
- control policy
- policy iteration
- control policies
- model free reinforcement learning
- learning algorithm
- inverse reinforcement learning
- model free
- partially observable domains
- transfer learning
- least squares
- approximate dynamic programming
- partially observable markov decision processes
- rotation invariant
- dynamic programming
- neural network
- policy gradient methods
- exploration exploitation tradeoff
- multi agent
- pose estimation
- decision problems
- optimal control
- rl algorithms
- infinite horizon
- machine learning
- natural actor critic
- agent receives
- approximate policy iteration
- control problems
- asymptotically optimal
- average reward
- gradient method
- state action
- temporal difference learning
- reinforcement learning methods
- continuous state