GS-LTS: 3D Gaussian Splatting-Based Adaptive Modeling for Long-Term Service Robots

TL;DR

GS-LTS tackles the challenge of maintaining accurate 3D scene representations for indoor service robots in dynamic environments. It couples a semantic-aware 3D Gaussian Splatting mapping with a Multi-Task Executor, a Change Detection Unit, and an Active Scene Updater to detect object changes from single egocentric views and edit the Gaussian scene accordingly. The authors also present a scalable simulation benchmark that generates object-level scene changes via configuration scripts, supporting systematic evaluation and sim-to-real transfer. Experimental results show GS-LTS achieves faster, higher-quality scene updates and improves reconstruction, navigation, and localization compared to image-only fine-tuning baselines, advancing 3DGS applicability for long-term robotics. Together, these contributions enable robust, dynamic scene understanding and manipulation for indoor service robots over extended time horizons.

Abstract

3D Gaussian Splatting (3DGS) has garnered significant attention in robotics for its explicit, high fidelity dense scene representation, demonstrating strong potential for robotic applications. However, 3DGS-based methods in robotics primarily focus on static scenes, with limited attention to the dynamic scene changes essential for long-term service robots. These robots demand sustained task execution and efficient scene updates-challenges current approaches fail to meet. To address these limitations, we propose GS-LTS (Gaussian Splatting for Long-Term Service), a 3DGS-based system enabling indoor robots to manage diverse tasks in dynamic environments over time. GS-LTS detects scene changes (e.g., object addition or removal) via single-image change detection, employs a rule-based policy to autonomously collect multi-view observations, and efficiently updates the scene representation through Gaussian editing. Additionally, we propose a simulation-based benchmark that automatically generates scene change data as compact configuration scripts, providing a standardized, user-friendly evaluation benchmark. Experimental results demonstrate GS-LTS's advantages in reconstruction, navigation, and superior scene updates-faster and higher quality than the image training baseline-advancing 3DGS for long-term robotic operations. Code and benchmark are available at: https://vipl-vsu.github.io/3DGS-LTS.

Figures (7)

• Figure 1: Three common types of scene changes in indoor scenes.
• Figure 2: System Overview. GS-LTS is a modular system designed for long-term service robots, which can adapt to object changes in the dynamic environments and update the 3DGS representation through periodic, automated operation of the Change Detection Unit and the Active Scene Updater.
• Figure 3: Impact of fine-tuning iterations on scene update quality.
• Figure 4: Rendering results after different fine-tuning iterations.
• Figure 5: 3D Localization Examples. Red bounding boxes indicate the results from GS-LTS (GT), while green ones from GS-LTS (CLIP).