Dext-Gen: Dexterous Grasping in Sparse Reward Environments with Full Orientation Control

TL;DR

Dext-Gen is a reinforcement learning framework enabling dexterous robotic grasping with full orientation control in sparse reward environments, demonstrating adaptability and unbiased policy learning across various scenarios.

Contribution

The paper introduces Dext-Gen, a novel RL framework that achieves full orientation control and unbiased grasping policies in sparse reward settings for diverse robotic hands.

Findings

Effective in simulated experiments across different scenarios

Achieves full orientation control of gripper and object

Provides reasonable training durations and flexibility to incorporate prior knowledge

Abstract

Reinforcement learning is a promising method for robotic grasping as it can learn effective reaching and grasping policies in difficult scenarios. However, achieving human-like manipulation capabilities with sophisticated robotic hands is challenging because of the problem's high dimensionality. Although remedies such as reward shaping or expert demonstrations can be employed to overcome this issue, they often lead to oversimplified and biased policies. We present Dext-Gen, a reinforcement learning framework for Dexterous Grasping in sparse reward ENvironments that is applicable to a variety of grippers and learns unbiased and intricate policies. Full orientation control of the gripper and object is achieved through smooth orientation representation. Our approach has reasonable training durations and provides the option to include desired prior knowledge. The effectiveness and adaptability of the framework to different scenarios is demonstrated in simulated experiments.