Offline Learning of Counterfactual Perception as Prediction for Real-World Robotic Reinforcement Learning.
Jun JinDaniel GravesCameron HaighJun LuoMartin JägersandPublished in: CoRR (2020)
Keyphrases
- reinforcement learning
- real world
- learning algorithm
- learning process
- learning agents
- prior knowledge
- learning capabilities
- online learning
- perception action
- robot control
- prediction accuracy
- supervised learning
- machine learning
- optimal policy
- learning problems
- wide range
- motor control
- data sets
- function approximation
- autonomous robots
- model free
- state space
- function approximators
- sensory data
- imitation learning
- eligibility traces