End-to-end Reinforcement Learning of Robotic Manipulation with Robust Keypoints Representation

TL;DR

This paper introduces an end-to-end reinforcement learning framework for robotic manipulation that uses a self-supervised keypoints representation to improve robustness and facilitate zero-shot transfer from simulation to real-world applications.

Contribution

The work presents a novel approach combining self-supervised keypoints learning with reinforcement learning for robotic manipulation, enabling robust and efficient sim-to-real transfer.

Findings

Effective manipulation in simulation and real-world scenarios

Robust keypoints improve learning efficiency and transferability

Zero-shot sim-to-real transfer achieved with domain randomization

Abstract

We present an end-to-end Reinforcement Learning(RL) framework for robotic manipulation tasks, using a robust and efficient keypoints representation. The proposed method learns keypoints from camera images as the state representation, through a self-supervised autoencoder architecture. The keypoints encode the geometric information, as well as the relationship of the tool and target in a compact representation to ensure efficient and robust learning. After keypoints learning, the RL step then learns the robot motion from the extracted keypoints state representation. The keypoints and RL learning processes are entirely done in the simulated environment. We demonstrate the effectiveness of the proposed method on robotic manipulation tasks including grasping and pushing, in different scenarios. We also investigate the generalization capability of the trained model. In addition to the robust keypoints representation, we further apply domain randomization and adversarial training examples to achieve zero-shot sim-to-real transfer in real-world robotic manipulation tasks.