Robot Generating Data for Learning Generalizable Visual Robotic Manipulation

It has been a popular trend in AI to pretrain foundation models on massive data. However, collecting sufficient offline training trajectories for robot learning is particularly expensive since valid control actions are required. Therefore, most existing robotic datasets are collected from human experts. We tackle such a data collection issue with a new framework called "robot self-teaching", which asks the robot to self-generate effective training data instead of relying on human demonstrators. Our key idea is to train a separate data-generation policy operating on the state space to automatically generate meaningful actions and trajectories with ever-growing complexities. Then, these generated data can be further used to train a visual policy with strong compositional generalization capabilities. We validate our framework in two visual manipulation testbeds, including a multi-object stacking domain and a popular RL benchmark "Franka kitchen". Experiments show that the final visual policy trained on self-generated data can accomplish novel testing goals that require long-horizon robot executions. Project website https://sites.google.com/view/robot-self-teaching.

Authors

• Yunfei Li
• Ying Yuan
• Jingzhi Cui
• Haoran Huan
• Wei Fu
• Jiaxuan Gao
• Zekai Xu
• Yi Wu

Keywords

• Training
• Visualization
• Stacking
• Robot control
• Training data
• Market research
• Robot learning
• Trajectory
• Complexity theory
• Robots
• State Space
• Foundation Model
• Final Policy
• Artificial Intelligence
• Challenging Task
• Data Generation
• Reachable
• Single Object
• Goal State
• Rating Task
• Building Structures
• Basic Tasks
• Policy Learning
• Universal Function
• Reinforcement Learning Agent
• Robot Trajectory
• Successional Trajectories
• Privileged Information