Industrial-Grade Sensor Simulation via Gaussian Splatting: A Modular Framework for Scalable Editing and Full-Stack Validation

2025-03-14

AI Summary

Existing NeRF-based sensor simulation methods suffer from poor industrial applicability, low efficiency, and insufficient scalability for large-scale autonomous driving validation. To address these limitations, this paper introduces the first industrial-grade multi-sensor simulation system built upon Gaussian Splatting (GS), enabling joint modeling of cameras and LiDAR. We propose an explicit Gaussian primitive library, design an editable and extensible scene editing pipeline, and integrate controllable diffusion models to ensure semantic consistency in scene generation and fusion. The system incorporates physics-aware sensor modeling and a modular architecture, achieving geometric and photometric fidelity while significantly reducing per-frame latency—enabling real-time, interpretable, explicit scene editing. Deployed within a full-stack traffic-dynamic simulation environment, the system effectively validates end-to-end algorithm robustness under diverse, realistic conditions.

Technology Category

• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Search and Optimization: Sampling/Simulation-based Search
• Computer Vision: Low Level & Physics-based Vision

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Sensor simulation is pivotal for scalable validation of autonomous driving systems, yet existing Neural Radiance Fields (NeRF) based methods face applicability and efficiency challenges in industrial workflows. This paper introduces a Gaussian Splatting (GS) based system to address these challenges: We first break down sensor simulator components and analyze the possible advantages of GS over NeRF. Then in practice, we refactor three crucial components through GS, to leverage its explicit scene representation and real-time rendering: (1) choosing the 2D neural Gaussian representation for physics-compliant scene and sensor modeling, (2) proposing a scene editing pipeline to leverage Gaussian primitives library for data augmentation, and (3) coupling a controllable diffusion model for scene expansion and harmonization. We implement this framework on a proprietary autonomous driving dataset supporting cameras and LiDAR sensors. We demonstrate through ablation studies that our approach reduces frame-wise simulation latency, achieves better geometric and photometric consistency, and enables interpretable explicit scene editing and expansion. Furthermore, we showcase how integrating such a GS-based sensor simulator with traffic and dynamic simulators enables full-stack testing of end-to-end autonomy algorithms. Our work provides both algorithmic insights and practical validation, establishing GS as a cornerstone for industrial-grade sensor simulation.

Problem

• Addresses challenges in industrial sensor simulation using Gaussian Splatting.
• Enables real-time rendering and explicit scene editing for autonomous systems.
• Improves simulation efficiency and consistency for full-stack autonomy testing.

Innovation

• Gaussian Splatting for real-time rendering
• 2D neural Gaussian for sensor modeling
• Controllable diffusion for scene expansion