Problem
Research questions and friction points this paper is trying to address.
autonomous driving
3D Gaussian Splatting
simulation
dynamic scene composition
closed-loop validation
DecoupleGS: Interactive 3D Gaussian Splatting for End-to-End Autonomous Driving Testing
🤖 AI Summary
Existing simulation environments struggle to simultaneously achieve high visual fidelity, strong interactivity, and real-time performance, thereby limiting closed-loop validation of end-to-end autonomous driving algorithms. This work proposes a decoupled 3D Gaussian splatting framework that separates scenes into a high-fidelity static background and controllable dynamic agents, leveraging an object-centric representation for efficient dynamic scene synthesis. Three key innovations—asset compression via perceptual pruning and vector quantization, map-guided geometric registration through semantic topological trajectory alignment, and agent-based relighting for environmental illumination transfer—effectively mitigate representation conflicts and enhance photometric consistency. The method significantly improves both metric accuracy and visual fidelity while maintaining high rendering efficiency, enabling a practical closed-loop sensor simulation platform for end-to-end autonomous driving.
📝 Abstract
End-to-end (E2E) autonomous driving algorithms require rigorous closed-loop validation in simulation environments offering high visual fidelity, strong interactivity, and real-time performance. Existing approaches, from game engines to static neural rendering, inherently trade off these requirements and struggle with the dynamic scene composition essential for E2E testing. To bridge this gap, we propose a novel decoupled 3D Gaussian Splatting (3DGS) framework tailored for large-scale E2E evaluation. We fundamentally decompose scenes into a high-fidelity static background and manipulable dynamic agents using an object-centric canonical representation. To resolve resulting representational conflicts, we introduce three targeted modules: (1) asset compression via perceptual pruning and vector quantization for real-time traffic rendering; (2) map-guided geometric registration leveraging semantic topology to strictly align trajectories; and (3) proxy-based relighting transferring ambient illumination for seamless photometric integration. Extensive experiments demonstrate that DecoupleGS achieves a balanced fidelity-efficiency trade-off, improves metric and photometric consistency, and provides a practical closed-loop sensor simulation platform for E2E autonomous driving evaluation.
Innovation
Methods, ideas, or system contributions that make the work stand out.
Decoupled 3D Gaussian Splatting
End-to-End Autonomous Driving
Dynamic Scene Composition
Real-Time Neural Rendering
Photometric Consistency
🔎 Similar Papers
No similar papers found.
💼 Related Jobs
