Learning Complex Dexterous Manipulation with Deep Reinforcement Learning and Demonstrations

TL;DR

This paper demonstrates that deep reinforcement learning, combined with human demonstrations, can effectively control high-dimensional dexterous hands for complex tasks, reducing sample complexity and improving robustness in simulated environments.

Contribution

It shows that model-free deep reinforcement learning can scale to high-dimensional dexterous manipulation and that demonstrations significantly enhance learning efficiency and policy robustness.

Findings

Successfully scaled DRL to 24-DoF hand manipulation

Reduced sample complexity with demonstrations to hours of robot experience

Produced natural and robust manipulation policies

Abstract

Dexterous multi-fingered hands are extremely versatile and provide a generic way to perform a multitude of tasks in human-centric environments. However, effectively controlling them remains challenging due to their high dimensionality and large number of potential contacts. Deep reinforcement learning (DRL) provides a model-agnostic approach to control complex dynamical systems, but has not been shown to scale to high-dimensional dexterous manipulation. Furthermore, deployment of DRL on physical systems remains challenging due to sample inefficiency. Consequently, the success of DRL in robotics has thus far been limited to simpler manipulators and tasks. In this work, we show that model-free DRL can effectively scale up to complex manipulation tasks with a high-dimensional 24-DoF hand, and solve them from scratch in simulated experiments. Furthermore, with the use of a small number of human demonstrations, the sample complexity can be significantly reduced, which enables learning with sample sizes equivalent to a few hours of robot experience. The use of demonstrations result in policies that exhibit very natural movements and, surprisingly, are also substantially more robust.