Interactive Imitation Learning in Robotics based on Simulations

TL;DR

This paper explores interactive imitation learning in robotics, implementing algorithms in simulations to compare with reinforcement learning, aiming to improve data efficiency and teaching ease.

Contribution

It introduces and evaluates IIL algorithms in multiple simulation scenarios, providing comprehensive analysis and comparison with RL methods.

Findings

IIL reduces data requirements compared to RL.

IIL performs effectively in complex, changing environments.

Extensive experiments demonstrate the advantages of IIL over traditional methods.

Abstract

The transformation towards intelligence in various industries is creating more demand for intelligent and flexible products. In the field of robotics, learning-based methods are increasingly being applied, with the purpose of training robots to learn to deal with complex and changing external environments through data. In this context, reinforcement learning and imitation learning are becoming research hotspots with their respective characteristics. However, the two have their own limitations in some cases, such as the high cost of data acquisition for reinforcement learning. Moreover, it is difficult for imitation learning to provide perfect demonstrations. As a branch of imitation learning, interactive imitation learning aims at transferring human knowledge to the agent through interactions between the demonstrator and the robot, which alleviates the difficulty of teaching. This thesis implements IIL algorithms in four simulation scenarios and conducts extensive experiments, aiming at providing exhaustive information about IIL methods both in action space and state space as well as comparison with RL methods.