Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Efficient Incremental SLAM via Information-Guided and Selective Optimization

Reza Arablouei

TL;DR

This paper addresses the challenge of achieving batch-level SLAM accuracy in a real-time incremental back-end. It introduces information-guided gating (IGG) to decide when to perform global optimization and selective partial optimization (SPO) to restrict GN iterations to the most affected variables, while preserving all measurements for global consistency. Theoretical analysis shows convergence to the same stationary point as full Gauss-Newton with the same local rates, and experiments demonstrate batch-level accuracy with substantial computation reductions, especially on large graphs. The approach enables scalable, real-time SLAM in data-rich environments and integrates smoothly with existing graph-based back-ends and libraries.

Abstract

We present an efficient incremental SLAM back-end that achieves the accuracy of full batch optimization while substantially reducing computational cost. The proposed approach combines two complementary ideas: information-guided gating (IGG) and selective partial optimization (SPO). IGG employs an information-theoretic criterion based on the log-determinant of the information matrix to quantify the contribution of new measurements, triggering global optimization only when a significant information gain is observed. This avoids unnecessary relinearization and factorization when incoming data provide little additional information. SPO executes multi-iteration Gauss-Newton (GN) updates but restricts each iteration to the subset of variables most affected by the new measurements, dynamically refining this active set until convergence. Together, these mechanisms retain all measurements to preserve global consistency while focusing computation on parts of the graph where it yields the greatest benefit. We provide theoretical analysis showing that the proposed approach maintains the convergence guarantees of full GN. Extensive experiments on benchmark SLAM datasets show that our approach consistently matches the estimation accuracy of batch solvers, while achieving significant computational savings compared to conventional incremental approaches. The results indicate that the proposed approach offers a principled balance between accuracy and efficiency, making it a robust and scalable solution for real-time operation in dynamic data-rich environments.

Efficient Incremental SLAM via Information-Guided and Selective Optimization

TL;DR

This paper addresses the challenge of achieving batch-level SLAM accuracy in a real-time incremental back-end. It introduces information-guided gating (IGG) to decide when to perform global optimization and selective partial optimization (SPO) to restrict GN iterations to the most affected variables, while preserving all measurements for global consistency. Theoretical analysis shows convergence to the same stationary point as full Gauss-Newton with the same local rates, and experiments demonstrate batch-level accuracy with substantial computation reductions, especially on large graphs. The approach enables scalable, real-time SLAM in data-rich environments and integrates smoothly with existing graph-based back-ends and libraries.

Abstract

We present an efficient incremental SLAM back-end that achieves the accuracy of full batch optimization while substantially reducing computational cost. The proposed approach combines two complementary ideas: information-guided gating (IGG) and selective partial optimization (SPO). IGG employs an information-theoretic criterion based on the log-determinant of the information matrix to quantify the contribution of new measurements, triggering global optimization only when a significant information gain is observed. This avoids unnecessary relinearization and factorization when incoming data provide little additional information. SPO executes multi-iteration Gauss-Newton (GN) updates but restricts each iteration to the subset of variables most affected by the new measurements, dynamically refining this active set until convergence. Together, these mechanisms retain all measurements to preserve global consistency while focusing computation on parts of the graph where it yields the greatest benefit. We provide theoretical analysis showing that the proposed approach maintains the convergence guarantees of full GN. Extensive experiments on benchmark SLAM datasets show that our approach consistently matches the estimation accuracy of batch solvers, while achieving significant computational savings compared to conventional incremental approaches. The results indicate that the proposed approach offers a principled balance between accuracy and efficiency, making it a robust and scalable solution for real-time operation in dynamic data-rich environments.
Paper Structure (19 sections, 5 theorems, 26 equations, 5 figures, 2 tables, 2 algorithms)

This paper contains 19 sections, 5 theorems, 26 equations, 5 figures, 2 tables, 2 algorithms.

Table of Contents

  1. Introduction
  2. Contributions
  3. Background
  4. Proposed Approach
  5. Graph Update and Initial Linearization
  6. Information-Guided Gating
  7. Selective Partial Solve
  8. Determining Affected Variables
  9. Selective Partial Optimization
  10. Computational Complexity
  11. Theoretical Analysis
  12. IGG with Localized Updates
  13. SPO as Block-Coordinate Descent
  14. Combined Convergence Guarantee
  15. Experiments
  16. ...and 4 more sections

Key Result

Lemma 4.1

Suppose $c(\mathbf{x})$ has a Lipschitz-continuous GN Hessian in a neighborhood of the solution $\mathbf{x}^\star$. Then, the updates generated under the IGG trigger, with localized solves on $\mathcal{S}'$ when $\Delta \eta_t < \tau_\eta$, produce a monotonically non-increasing cost sequence and co

Figures (5)

  • Figure 1: ATE over increments for four datasets.
  • Figure 2: Cumulative update FLOPs over increments for four dataset.
  • Figure 3: Information gain over increments for four dataset.
  • Figure 4: Performance of the proposed algorithm on the MIT dataset for different values of the update threshold $\tau_{d}$.
  • Figure 5: Performance of the proposed approach on the FR079 dataset for different values of the information-gain threshold $\tau_{\eta}$.

Theorems & Definitions (8)

  • Lemma 4.1
  • proof
  • Corollary 4.2
  • proof
  • Lemma 4.3
  • proof
  • Theorem 4.4
  • Corollary 4.5