Transferring policy of deep reinforcement learning from simulation to reality for robotics.
Hao JuRongshun JuanRandy GomezKeisuke NakamuraGuangliang LiPublished in: Nat. Mac. Intell. (2022)
Keyphrases
- reinforcement learning
- optimal policy
- real robot
- action selection
- markov decision process
- control policy
- transfer learning
- reinforcement learning problems
- reinforcement learning algorithms
- robot control
- policy search
- markov decision processes
- function approximation
- state space
- partially observable environments
- computer vision
- artificial intelligence
- policy gradient
- action space
- function approximators
- machine intelligence
- approximate dynamic programming
- simulation model
- simulation environment
- continuous state
- policy evaluation
- partially observable domains
- control policies
- state action
- machine learning
- model free
- finite state
- robotic systems
- decision problems
- active learning
- learning algorithm