Learning-Based Gaussian Belief Propagation for Bundle Adjustment in Visual SLAM

Yu Siang Feng*, Jian Yu Chen, Han Chun Wang, Chih Wei Huang, Jann Long Chern

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference contribution


Bundle adjustment (BA) is the major optimization step simultaneously refining 3D coordinates and accounts for a large portion of execution time in visual simultaneous localization and mapping (SLAM). While the Levenberg-Marquardt (LM) based algorithms have been commonly used for fast BA, recent solutions adopting iterative and originally slow Gaussian belief propagation (GBP) show its potential to be fast and accurate on emerging computation platforms. We propose a novel architecture to predict the message passing in GBP with deep neural networks. The model generates messages several iterations ahead to significantly reduce the number of required computation loops. Also, the process converges with hyperparameter tuning and avoids the dependency of an arbitrary damping factor for GBP to be stabilized. Compared with standard GBP, the learning-based approach achieves the same level of accuracy while running 17.7 times faster under GPU acceleration.

Original languageEnglish
Title of host publication2022 IEEE GLOBECOM Workshops, GC Wkshps 2022 - Proceedings
PublisherInstitute of Electrical and Electronics Engineers Inc.
Number of pages6
ISBN (Electronic)9781665459754
Publication statusPublished - 2022
Event2022 IEEE GLOBECOM Workshops, GC Wkshps 2022 - Virtual, Online, Brazil
Duration: 2022 Dec 42022 Dec 8

Publication series

Name2022 IEEE GLOBECOM Workshops, GC Wkshps 2022 - Proceedings


Conference2022 IEEE GLOBECOM Workshops, GC Wkshps 2022
CityVirtual, Online


  • Bundle adjustment
  • Deep learning
  • Gaussian Belief Propagation
  • Machine learning
  • Neural network
  • Spatial AI
  • Visual SLAM

ASJC Scopus subject areas

  • Instrumentation
  • Computer Networks and Communications
  • Control and Optimization


Dive into the research topics of 'Learning-Based Gaussian Belief Propagation for Bundle Adjustment in Visual SLAM'. Together they form a unique fingerprint.

Cite this