Learning Synergies between Pushing and Grasping with Self-supervised Deep Reinforcement Learning

TL;DR

This paper presents a self-supervised deep reinforcement learning approach for robotic manipulation that learns to synergize pushing and grasping actions, improving efficiency and success rates in cluttered environments.

Contribution

It introduces a joint learning framework for pushing and grasping using two convolutional networks trained via Q-learning, enabling robots to discover complex manipulation strategies from scratch.

Findings

Learned pushing actions facilitate better grasping in cluttered scenes.

Achieved higher grasp success rates compared to baseline methods.

System generalizes effectively to novel objects.

Abstract

Skilled robotic manipulation benefits from complex synergies between non-prehensile (e.g. pushing) and prehensile (e.g. grasping) actions: pushing can help rearrange cluttered objects to make space for arms and fingers; likewise, grasping can help displace objects to make pushing movements more precise and collision-free. In this work, we demonstrate that it is possible to discover and learn these synergies from scratch through model-free deep reinforcement learning. Our method involves training two fully convolutional networks that map from visual observations to actions: one infers the utility of pushes for a dense pixel-wise sampling of end effector orientations and locations, while the other does the same for grasping. Both networks are trained jointly in a Q-learning framework and are entirely self-supervised by trial and error, where rewards are provided from successful grasps. In this way, our policy learns pushing motions that enable future grasps, while learning grasps that can leverage past pushes. During picking experiments in both simulation and real-world scenarios, we find that our system quickly learns complex behaviors amid challenging cases of clutter, and achieves better grasping success rates and picking efficiencies than baseline alternatives after only a few hours of training. We further demonstrate that our method is capable of generalizing to novel objects. Qualitative results (videos), code, pre-trained models, and simulation environments are available at http://vpg.cs.princeton.edu