Reinforcement Learning Based Pushing and Grasping Objects from Ungraspable Poses

TL;DR

This paper presents a model-free deep reinforcement learning approach that enables robots to push and grasp objects from ungraspable poses by combining pushing and grasping actions, improving efficiency and generalization.

Contribution

The authors develop a unified RL framework with a shared policy network for pushing and grasping, trained efficiently and transferable to real robots without fine-tuning.

Findings

Faster convergence than separate networks (2.5x).

Good generalization to unseen objects with complex shapes.

Successful real robot transfer using domain adaptation.

Abstract

Grasping an object when it is in an ungraspable pose is a challenging task, such as books or other large flat objects placed horizontally on a table. Inspired by human manipulation, we address this problem by pushing the object to the edge of the table and then grasping it from the hanging part. In this paper, we develop a model-free Deep Reinforcement Learning framework to synergize pushing and grasping actions. We first pre-train a Variational Autoencoder to extract high-dimensional features of input scenario images. One Proximal Policy Optimization algorithm with the common reward and sharing layers of Actor-Critic is employed to learn both pushing and grasping actions with high data efficiency. Experiments show that our one network policy can converge 2.5 times faster than the policy using two parallel networks. Moreover, the experiments on unseen objects show that our policy can generalize to the challenging case of objects with curved surfaces and off-center irregularly shaped objects. Lastly, our policy can be transferred to a real robot without fine-tuning by using CycleGAN for domain adaption and outperforms the push-to-wall baseline.