Robotic Offline RL from Internet Videos via Value-Function Pre-Training.
Chethan BhatejaDerek GuoDibya GhoshAnikait SinghManan TomarQuan VuongYevgen ChebotarSergey LevineAviral KumarPublished in: CoRR (2023)
Keyphrases
- reinforcement learning
- video sequences
- real time
- training set
- robotic systems
- neural network
- multi agent
- optimal policy
- training process
- training samples
- learning algorithm
- video frames
- human activities
- spatio temporal
- optimal control
- video database
- internet users
- function approximators
- surgical training
- fully labeled