RoboArmGS: High-Quality Robotic Arm Splatting via B\'ezier Curve Refinement

TL;DR

RoboArmGS introduces a hybrid representation that refines URDF-based robotic arm models with learnable Bézier curves, significantly improving the realism and accuracy of 3D Gaussian rendering of robotic motion from real observations.

Contribution

The paper presents RoboArmGS, a novel method combining URDF models with learnable Bézier curves for better motion modeling and rendering of robotic arms.

Findings

Achieves state-of-the-art motion modeling accuracy.

Produces photorealistic rendering of robotic arms.

Provides a new dataset RoboArm4D for evaluation.

Abstract

Constructing photorealistic and controllable robotic arm digital assets from real observations is fundamental to robotic applications. Current approaches naively bind static 3D Gaussians according to URDF links, forcing them to follow an URDF-rigged motion passively. However, the idealized URDF-rigged motion cannot accurately model the actual motion captured in real-world observations, leading to severe rendering artifacts in 3D Gaussians. To address these challenges, we propose RoboArmGS, a novel hybrid representation that refines the URDF-rigged motion with learnable B\'ezier curves, enabling more accurate real-world motion modeling. To be more specific, we present a learnable B\'ezier Curve motion refiner that corrects per-joint residuals to address mismatches between real-world motion and URDF-rigged motion. RoboArmGS enables the learning of more accurate real-world motion while achieving a coherent binding of 3D Gaussians across arm parts. To support future research, we contribute a carefully collected dataset named RoboArm4D, which comprises several widely used robotic arms for evaluating the quality of building high-quality digital assets. We evaluate our approach on RoboArm4D, and RoboArmGS achieves state-of-the-art performance in real-world motion modeling and rendering quality. The code and dataset will be released.