Data Scaling Laws in Imitation Learning for Robotic Manipulation

TL;DR

This study explores how data scaling influences the generalization of imitation learning policies in robotic manipulation, revealing power-law relationships and emphasizing environment and object diversity over sheer data volume.

Contribution

It provides the first comprehensive empirical analysis of data scaling laws in robotic imitation learning, demonstrating how diversity impacts generalization and proposing an efficient data collection strategy.

Findings

Generalization follows a power-law with data quantity.

Diversity of environments and objects is crucial.

Limited demonstrations per environment suffice for high success rates.

Abstract

Data scaling has revolutionized fields like natural language processing and computer vision, providing models with remarkable generalization capabilities. In this paper, we investigate whether similar data scaling laws exist in robotics, particularly in robotic manipulation, and whether appropriate data scaling can yield single-task robot policies that can be deployed zero-shot for any object within the same category in any environment. To this end, we conduct a comprehensive empirical study on data scaling in imitation learning. By collecting data across numerous environments and objects, we study how a policy's generalization performance changes with the number of training environments, objects, and demonstrations. Throughout our research, we collect over 40,000 demonstrations and execute more than 15,000 real-world robot rollouts under a rigorous evaluation protocol. Our findings reveal several intriguing results: the generalization performance of the policy follows a roughly power-law relationship with the number of environments and objects. The diversity of environments and objects is far more important than the absolute number of demonstrations; once the number of demonstrations per environment or object reaches a certain threshold, additional demonstrations have minimal effect. Based on these insights, we propose an efficient data collection strategy. With four data collectors working for one afternoon, we collect sufficient data to enable the policies for two tasks to achieve approximately 90% success rates in novel environments with unseen objects.