Reward Shaping to Learn Natural Object Manipulation With an Anthropomorphic Robotic Hand and Hand Pose Priors via On-Policy Reinforcement Learning.
Patricio RiveraJiheon OhEdwin ValarezoGahyeon RyuHwanseok JungJin Hyunk LeeJin Gyun JeongTae-Seong KimPublished in: ICTC (2021)
Keyphrases
- object manipulation
- hand pose
- reward shaping
- reinforcement learning
- agent learns
- robot control
- complex domains
- pose estimation
- human hand
- markov decision problems
- function approximators
- position and orientation
- optimal policy
- degrees of freedom
- hand tracking
- reinforcement learning algorithms
- policy search
- reward function
- learning agent
- function approximation
- mobile robot
- manipulation tasks
- robotic systems
- state space
- robot arm
- action selection
- markov decision process
- markov decision processes
- partially observable
- dynamic programming
- temporal difference
- action space
- real robot
- hand postures
- autonomous robots
- learning algorithm
- robotic arm
- d objects
- feature points
- skin color
- real time
- computer vision
- hand gestures
- machine learning