Meta-Reinforcement Learning for Robotic Industrial Insertion Tasks

Robotic insertion tasks are characterized by contact and friction mechanics, making them challenging for conventional feedback control methods due to unmodeled physical effects. Reinforcement learning (RL) is a promising approach for learning control policies in such settings. However, RL can be unsafe during exploration and might require a large amount of real-world training data, which is expensive to collect. In this paper, we study how to use meta-reinforcement learning to solve the bulk of the problem in simulation by solving a family of simulated industrial insertion tasks and then adapt policies quickly in the real world. We demonstrate our approach by training an agent to successfully perform challenging real-world insertion tasks using less than 20 trials of real-world experience.

Authors

• Gerrit Schoettler
• Ashvin Nair
• Juan Aparicio Ojea
• Sergey Levine
• Eugen Solowjow

Keywords

• Connectors
• Training
• Adaptation models
• Service robots
• Training data
• Reinforcement learning
• Task analysis
• Robotic Tasks
• Industrial Tasks
• Insertion Task
• Meta Reinforcement Learning
• Challenging Task
• Feedback Control
• Real-world Tasks
• Simulated Task
• Training Time
• Difficult Task
• Control Parameters
• End Of The Trial
• Simulation Environment
• Depth Images
• Sampling Efficiency
• Vertical Force
• Reward Function
• Deep Reinforcement Learning
• End-effector
• Random Search
• Policy Learning
• Simulation Training
• Stochastic Policy
• Robot Manipulator
• Deterministic Policy
• End-effector Position
• Impedance Control
• Pattern Search
• Distribution Of Tasks
• Robot Control