DROID: Minimizing the Reality Gap using Single-Shot Human Demonstration

TL;DR

This paper introduces DROID, a framework that uses a single human demonstration to identify and optimize the simulator's dynamics parameters, improving the transfer of policies from simulation to real-world tasks like door opening.

Contribution

DROID is a novel method that leverages single-shot human demonstrations to better match simulation dynamics with reality, enhancing policy transfer and generalization.

Findings

DROID accurately identifies the dynamics discrepancy between simulation and real world.

Optimizing simulator parameters improves policy transfer success.

The method generalizes to related tasks beyond door opening.

Abstract

Reinforcement learning (RL) has demonstrated great success in the past several years. However, most of the scenarios focus on simulated environments. One of the main challenges of transferring the policy learned in a simulated environment to real world, is the discrepancy between the dynamics of the two environments. In prior works, Domain Randomization (DR) has been used to address the reality gap for both robotic locomotion and manipulation tasks. In this paper, we propose Domain Randomization Optimization IDentification (DROID), a novel framework to exploit single-shot human demonstration for identifying the simulator's distribution of dynamics parameters, and apply it to training a policy on a door opening task. Our results show that the proposed framework can identify the difference in dynamics between the simulated and the real worlds, and thus improve policy transfer by optimizing the simulator's randomization ranges. We further illustrate that based on these same identified parameters, our method can generalize the learned policy to different but related tasks.