Sample-efficient Adversarial Imitation Learning from Observation

TL;DR

This paper introduces a sample-efficient adversarial imitation learning algorithm that leverages trajectory-centric reinforcement learning ideas, enabling effective imitation from observation with fewer demonstrations, demonstrated on a robot arm in simulation and reality.

Contribution

The paper proposes a novel algorithm that reduces sample complexity in adversarial imitation learning by integrating trajectory-centric reinforcement learning techniques.

Findings

Improved learning rate and efficiency in imitation tasks.

Successful application on physical robot arm and simulation.

Reduced number of demonstrations needed for effective imitation.

Abstract

Imitation from observation is the framework of learning tasks by observing demonstrated state-only trajectories. Recently, adversarial approaches have achieved significant performance improvements over other methods for imitating complex behaviors. However, these adversarial imitation algorithms often require many demonstration examples and learning iterations to produce a policy that is successful at imitating a demonstrator's behavior. This high sample complexity often prohibits these algorithms from being deployed on physical robots. In this paper, we propose an algorithm that addresses the sample inefficiency problem by utilizing ideas from trajectory centric reinforcement learning algorithms. We test our algorithm and conduct experiments using an imitation task on a physical robot arm and its simulated version in Gazebo and will show the improvement in learning rate and efficiency.