QT-Opt: Scalable Deep Reinforcement Learning for Vision-Based Robotic Manipulation.
Dmitry KalashnikovAlex IrpanPeter PastorJulian IbarzAlexander HerzogEric JangDeirdre QuillenEthan HollyMrinal KalakrishnanVincent VanhouckeSergey LevinePublished in: CoRR (2018)
Keyphrases
- reinforcement learning
- manipulation tasks
- real robot
- real time
- human computer interaction
- function approximation
- reinforcement learning algorithms
- mobile robot
- perception action
- vision system
- robot control
- highly scalable
- learning algorithm
- model free
- markov decision processes
- parallel robot
- robotic systems
- imitation learning
- optimal control
- machine learning
- transfer learning
- robot navigation
- action selection
- temporal difference
- augmented reality
- lightweight
- state space
- object manipulation
- computer vision
- multi agent
- encoding scheme
- genetic algorithm
- robotic control
- driver assistance systems
- service robots
- temporal difference learning
- learning problems
- optimal policy
- motion planning