RoboNet: Large-Scale Multi-Robot Learning

Abstract

Robot learning has emerged as a promising tool for taming the complexity and diversity of the real world. Methods based on high-capacity models, such as deep networks, hold the promise of providing effective generalization to a wide range of open-world environments. However, these same methods typically require large amounts of diverse training data to generalize effectively. In contrast, most robotic learning experiments are small-scale, single-domain, and single-robot. This leads to a frequent tension in robotic learning: how can we learn generalizable robotic controllers without having to collect impractically large amounts of data for each separate experiment? In this paper, we propose RoboNet, an open database for sharing robotic experience, which provides an initial pool of 15 million video frames, from 7 different robot platforms, and study how it can be used to learn generalizable models for vision-based robotic manipulation. We combine the dataset with two different learning algorithms: visual foresight, which uses forward video prediction models, and supervised inverse models. Our experiments test the learned algorithms' ability to work across new objects, new tasks, new scenes, new camera viewpoints, new grippers, or even entirely new robots. In our final experiment, we find that by pre-training on RoboNet and fine-tuning on data from a held-out Franka or Kuka robot, we can exceed the performance of a robot-specific training approach that uses 4x-20x more data. For videos and data, see the project webpage: https://www.robonet.wiki/

accepted at the Conference on Robot Learning (CoRL) 2019

References in corpus (24)

• Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networks
• Gaussian Processes for Data-Efficient Learning in Robotics and Control
• An Empirical Evaluation of Deep Learning on Highway Driving
• One-Shot Visual Imitation Learning via Meta-Learning
• Learning Dexterous In-Hand Manipulation
• Visual Foresight: Model-Based Deep Reinforcement Learning for Vision-Based Robotic Control
• Exploration by Random Network Distillation
• Stochastic Adversarial Video Prediction
• Transfer from Simulation to Real World through Learning Deep Inverse Dynamics Model
• Self-Supervised Visual Planning with Temporal Skip Connections
• Learning Invariant Feature Spaces to Transfer Skills with Reinforcement Learning
• ChauffeurNet: Learning to Drive by Imitating the Best and Synthesizing the Worst
• Combining Model-Based and Model-Free Updates for Trajectory-Centric Reinforcement Learning
• Model Predictive Path Integral Control using Covariance Variable Importance Sampling
• Robot Learning in Homes: Improving Generalization and Reducing Dataset Bias
• RoboTurk: A Crowdsourcing Platform for Robotic Skill Learning through Imitation
• Multiple Interactions Made Easy (MIME): Large Scale Demonstrations Data for Imitation
• Modular meta-learning
• Manipulation by Feel: Touch-Based Control with Deep Predictive Models
• Learning Latent Plans from Play
• Self-Supervised Exploration via Disagreement
• Improvisation through Physical Understanding: Using Novel Objects as Tools with Visual Foresight
• Scaling Autoregressive Video Models
• Time Reversal as Self-Supervision

Cited by in corpus (9)

• Offline Reinforcement Learning: Tutorial, Review, and Perspectives on Open Problems
• A Review on Deep Learning Techniques for Video Prediction
• MoVi: A Large Multipurpose Motion and Video Dataset
• ManiSkill: Generalizable Manipulation Skill Benchmark with Large-Scale Demonstrations
• Data-efficient visuomotor policy training using reinforcement learning and generative models
• S4RL: Surprisingly Simple Self-Supervision for Offline Reinforcement Learning
• Weakly-Supervised Reinforcement Learning for Controllable Behavior
• Transfer Reinforcement Learning across Homotopy Classes
• Episodic Memory Model for Learning Robotic Manipulation Tasks