Proactive Action Visual Residual Reinforcement Learning for Contact-Rich Tasks Using a Torque-Controlled Robot.
Yunlei ShiZhaopeng ChenHongxu LiuSebastian RiedelChunhui GaoQian FengJun DengJianwei ZhangPublished in: CoRR (2020)
Keyphrases
- reinforcement learning
- sensory inputs
- real robot
- robotic systems
- action selection
- contact force
- mobile robot
- robot control
- autonomous robots
- visual tasks
- action selection mechanism
- transfer learning
- high level
- partially observable domains
- landmark recognition
- manipulation tasks
- computer controlled
- visual input
- learning algorithm
- agent learns
- service robots
- action space
- robot navigation
- markov decision processes
- feedback loop
- human users
- human robot interaction
- multi robot
- robot arm
- visual features
- vision system
- temporal difference
- force feedback
- state action
- multi agent
- model free
- position and orientation
- reward shaping
- function approximation
- visual landmarks
- reward signal
- primitive actions