uncertainty-aware factor graph optimization

This work addresses the limitations of traditional factor graph estimation, which relies on local optimization and is prone to suboptimal solutions, thereby failing to meet the reliability demands of safety-critical applications requiring global optimality. While existing certifiable methods guarantee global optimality, their high computational complexity hinders practical deployment. The paper presents the first insight that the structure of factor graphs remains invariant under both Shor’s convex relaxation and Burer–Monteiro low-rank factorization. Leveraging this property, the authors develop a certifiable estimation framework compatible with mainstream factor graph libraries. By integrating QCQP modeling, Riemannian Staircase optimization, and native factor graph structure, the approach delivers verifiably globally optimal state estimates without altering existing workflows, significantly enhancing the practicality and deployability of certifiable estimation in real-world systems.

Existing LiDAR SLAM methods suffer from limited robustness in joint pose estimation and scene reconstruction—particularly under degenerate geometries and low-quality point clouds. To address this, we propose a surfel-based LiDAR Bundle Adjustment (LiDAR BA) framework. This work introduces surfels into LiDAR BA for the first time, enabling scene-adaptive generalized uncertainty modeling that jointly encodes geometric degeneracy and measurement confidence. Our method integrates probabilistic residual weighting, point-cloud geometric consistency constraints, and Ceres-based nonlinear optimization to achieve end-to-end joint refinement of poses and surfel maps. Evaluated on multiple public benchmarks, the framework significantly improves localization accuracy and mapping robustness across challenging scenarios. Furthermore, the open-source implementation supports real-time execution on embedded platforms.

This work addresses the challenge in factor graph–based state estimation where local optimization methods often converge to suboptimal solutions, while existing globally optimal approaches based on convex relaxation are hindered by modeling complexity and high computational cost. To overcome these limitations, the paper proposes an efficient method for globally optimal estimation that automatically constructs a semidefinite programming (SDP)–based convex relaxation within general factor graphs. The approach innovatively leverages the Bayes tree structure from the GTSAM framework together with chordal sparsity to decompose and solve the SDP problem efficiently. Experimental results on 3D pose-graph SLAM and 2D localization benchmarks demonstrate that the proposed method achieves global optimality while significantly outperforming conventional local solvers in both scalability and computational efficiency.

This work addresses the inconsistency in state estimation and computational inefficiency arising from the rigid structure of factor graphs under asynchronous multi-sensor measurements. To overcome these limitations, the authors propose an incremental dynamic factor graph construction method that incorporates an external evaluation criterion to select the optimal graph topology in real time. This approach enables synchronous fusion of asynchronous multi-source sensor data and supports on-the-fly graph compression to reduce the number of optimization variables. Experimental results demonstrate that the proposed method maintains map accuracy comparable to conventional approaches while reducing the average number of graph nodes by approximately 30%, thereby significantly lowering computational complexity. The key innovation lies in the evaluation-driven dynamic topology construction and compression mechanism, which effectively balances estimation accuracy and computational efficiency.

This work addresses the challenge of simultaneously mitigating long-range drift and preserving local geometric accuracy in multi-session UAV mapping. The authors propose an uncertainty-aware, coarse-to-fine map fusion and optimization framework that leverages RTK observations for spatiotemporal alignment to handle time offsets and missing data. The approach integrates scene-graph-based initial registration, dynamic time warping (DTW), and multi-output Gaussian processes (MOGP) to model trajectory uncertainty, and formulates a unified factor graph incorporating planar constraints for iterative refinement. Experimental results on real-world datasets demonstrate that the proposed method significantly improves both global consistency and local geometric fidelity of multi-session point cloud maps.

This study addresses the limitation of existing 6D pose tracking methods that rely on costly initialization or real-time reconstruction, thereby struggling to meet the real-time demands of robotic manipulation and augmented reality. To this end, this work proposes a lightweight framework for long-term rigid object tracking. Departing from conventional point-based optimization paradigms, the proposed method constructs a compact pose graph modeled exclusively with relative pose constraints weighted by geometrically aligned uncertainties, effectively decoupling computational complexity from the number of correspondences. Evaluated across four real-world benchmarks, the approach achieves accuracy comparable to reconstruction-based trackers at minimal optimization cost, offering an efficient and robust solution for real-time applications.

This study addresses the challenge of uncertainty-aware planning based on point cloud observations in large-scale outdoor digital twins by proposing the Informed BLT* algorithm. This method extends RRT* into belief space and introduces the Wasserstein metric to enable efficient node connectivity and information reuse, thereby avoiding redundant observation propagation. By integrating a Gaussian belief model with sampling-based planning, it establishes a framework for generating semantic digital twins. Experimental results demonstrate that the proposed algorithm significantly accelerates initial solution discovery while achieving competitive cost convergence, offering an efficient solution for uncertainty-aware planning in complex environments.

该研究解决了3D场景图中不确定性表示和传播问题，通过引入概率场景图（PSG）及高斯层次图（HGG），实现了实时感知与精确定位。

This work addresses the high communication overhead and low data efficiency in multi-robot collaborative SLAM, which often stems from reliance on low-level feature matching. The authors propose a distributed SLAM framework based on scene graph matching that leverages RGB-LiDAR fused point clouds for semantic segmentation, extracting discrete objects and their boundaries to construct scene graphs. Notably, inter-robot loop closures are achieved solely by exchanging object labels and centroids, eliminating dependence on raw feature descriptors. Integrated with a multi-stage communication scheme and distributed optimization, the method significantly reduces communication load while preserving localization and mapping accuracy, as demonstrated in both simulated and real-world experiments with legged robots across indoor and outdoor environments.

本文提出Scalix，一种实时单目SLAM框架，通过将学习到的深度线索整合进概率图模型中解决尺度不确定性问题，提高尺度一致性。