MISO: Multiresolution Submap Optimization for Efficient Globally Consistent Neural Implicit Reconstruction

2025-04-27

AI Summary

To address the low optimization efficiency and poor global consistency in neural implicit SLAM backends for large-scale environments, this paper proposes a multi-resolution subgraph hierarchical optimization framework. Methodologically: (1) a hierarchical subgraph structure is constructed to jointly enable local high-precision refinement and global coarse-grained alignment; (2) a learning-driven implicit feature initialization scheme is introduced to bypass full-scene signed distance function (SDF) decoding; and (3) a differentiable subgraph alignment module and multi-scale feature fusion mechanism are designed to enhance cross-scale consistency. Experiments on large real-world benchmarks demonstrate that our approach significantly improves computational efficiency—achieving a 2.3× speedup over baseline methods—and reconstruction accuracy—reducing Chamfer distance by 37%. To the best of our knowledge, this is the first method to achieve efficient, scalable, and globally consistent neural implicit 3D reconstruction.

Technology Category

• Search and Optimization: Learning to Search
• Computer Vision: Learning & Optimization for CV
• Machine Learning: Optimization

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
• Semantics and Knowledge: Scalable techniques for the creation, curation, publication, maintenance, and consumption of large, Web-based, structured, reusable, knowledge graphs and ontologies

Abstract

Neural implicit representations have had a significant impact on simultaneous localization and mapping (SLAM) by enabling robots to build continuous, differentiable, and high-fidelity 3D maps from sensor data. However, as the scale and complexity of the environment increase, neural SLAM approaches face renewed challenges in the back-end optimization process to keep up with runtime requirements and maintain global consistency. We introduce MISO, a hierarchical optimization approach that leverages multiresolution submaps to achieve efficient and scalable neural implicit reconstruction. For local SLAM within each submap, we develop a hierarchical optimization scheme with learned initialization that substantially reduces the time needed to optimize the implicit submap features. To correct estimation drift globally, we develop a hierarchical method to align and fuse the multiresolution submaps, leading to substantial acceleration by avoiding the need to decode the full scene geometry. MISO significantly improves computational efficiency and estimation accuracy of neural signed distance function (SDF) SLAM on large-scale real-world benchmarks.

Problem

• Efficient globally consistent neural implicit reconstruction
• Scalable optimization for large-scale neural SLAM
• Hierarchical submap alignment to reduce computational cost

Innovation

• Hierarchical optimization with multiresolution submaps
• Learned initialization for faster implicit submap optimization
• Hierarchical alignment and fusion for global consistency