ActiveGS: Active Scene Reconstruction using Gaussian Splatting

2024-12-23

IEEE Robotics and Automation Letters

AI Summary

To address the challenge of active 3D reconstruction in unknown environments using RGB-D cameras mounted on mobile robots (e.g., UAVs), this paper proposes a hybrid mapping framework integrating Gaussian splatting with voxel-based maps. Methodologically, it jointly optimizes a high-fidelity Gaussian splat map and a coarse-grained voxel hash map, incorporates confidence-driven uncertainty modeling to identify under-reconstructed regions, and enables closed-loop perception–planning co-optimization for active exploration and collision-free path planning leveraging both geometric and semantic voxel-space information. The core innovations include the first bidirectional coupling mechanism between Gaussian splats and voxel maps, and a confidence-guided active SLAM strategy. Evaluated on a real UAV platform, the method achieves significantly improved reconstruction completeness and geometric accuracy, outperforming state-of-the-art approaches in both exploration efficiency and reconstruction quality.

Technology Category

• Intelligent Robots: Localization, Mapping, and Navigation
• Planning, Routing, and Scheduling: Replanning and Plan Repair
• Search and Optimization: Mixed Discrete/Continuous Search

Application Category

• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Robotics applications often rely on scene reconstructions to enable downstream tasks. In this work, we tackle the challenge of actively building an accurate map of an unknown scene using an RGB-D camera on a mobile platform. We propose a hybrid map representation that combines a Gaussian splatting map with a coarse voxel map, leveraging the strengths of both representations: the high-fidelity scene reconstruction capabilities of Gaussian splatting and the spatial modelling strengths of the voxel map. At the core of our framework is an effective confidence modelling technique for the Gaussian splatting map to identify under-reconstructed areas, while utilising spatial information from the voxel map to target unexplored areas and assist in collision-free path planning. By actively collecting scene information in under-reconstructed and unexplored areas for map updates, our approach achieves superior Gaussian splatting reconstruction results compared to state-of-the-art approaches. Additionally, we demonstrate the real-world applicability of our framework using an unmanned aerial vehicle.

Problem

• Actively reconstructing unknown scenes using RGB-D camera
• Combining Gaussian splatting and voxel maps for accurate modeling
• Improving reconstruction by targeting under-reconstructed and unexplored areas

Innovation

• Hybrid map combines Gaussian splatting and voxel
• Confidence modeling identifies under-reconstructed areas
• Active data collection improves reconstruction accuracy