A sensorimotor reinforcement learning framework for physical Human-Robot Interaction

Modeling of physical human-robot collaborations is generally a challenging problem due to the unpredictive nature of human behavior. To address this issue, we present a data-efficient reinforcement learning framework which enables a robot to learn how to collaborate with a human partner. The robot learns the task from its own sensorimotor experiences in an unsupervised manner.

Authors

• Ali Ghadirzadeh
• Judith Bütepage
• Atsuto Maki
• Danica Kragic
• Mårten Björkman

Keywords

• Robot sensing systems
• Predictive models
• Trajectory
• Impedance
• Learning (artificial intelligence)
• Uncertainty
• Physical Interaction
• Physical Human-robot Interaction
• Human Behavior
• Gaussian Process
• Forward Model
• Bayesian Optimization
• Optimal Action
• Gaussian Process Model
• Human Partner
• Action-value Function
• Human-robot Collaboration
• Ball Position
• Objective Function
• Exponential Function
• Angular Velocity
• Position Error
• Kriging
• Interaction Forces
• Propagation Model
• Optimal Policy
• Ball Velocity
• Impedance Control
• Path Planning
• Human Motion
• Covariance Function
• Proactive Behavior
• Revolute Joints
• State-action Pair
• Upper Confidence Bound
• Visual Servoing