SplatSim: Zero-Shot Sim2Real Transfer of RGB Manipulation Policies Using Gaussian Splatting

Mohammad Nomaan Qureshi, Sparsh Garg, Francisco Yandun, David Held, George Kantor, Abhisesh Silwal

Keywords: Robotics data, Sim2Real, Manipulation Policy Learning

TL;DR: Photorealistic data generation for manipulation policies using Gaussian Splatting.

Abstract: Sim2Real transfer, particularly for manipulation policies relying on RGB images, remains a critical challenge in robotics due to the significant domain shift between synthetic and real-world visual data. In this paper, we propose \textit{SplatSim}, a novel framework that leverages Gaussian Splatting as the primary rendering primitive to reduce the Sim2Real gap for RGB-based manipulation policies. By replacing traditional mesh representations with Gaussian Splats in simulators, \textit{SplatSim} produces highly photorealistic synthetic data while maintaining the scalability and cost-efficiency of simulation. We demonstrate the effectiveness of our framework by training manipulation policies within \textit{SplatSim} and deploying them in the real world in a zero-shot manner, achieving an average success rate of 86.25\%, compared to 97.5\% for policies trained on real-world data. Videos can be found on our project page: https://splatsim.github.io

Submission Number: 29