Distributed Pose Graph Optimization Via Contractive Belief Sharing
Xiangyu Liu, Margarita Chli
AI summary
Problem
Current distributed pose graph optimization methods either require excessive communication rounds for reliable convergence or risk divergence on loopy, noisy graphs. There is a need for a scalable, fully distributed solver that guarantees stability without centralized synchronization.
Approach
The authors propose a two-stage message-passing algorithm that integrates local MAP optimization with neighbor-only belief sharing, using a Hellinger-distance-based damping rule to contract updates and ensure convergence.
Key results
- First DPGO method to regulate step size via belief divergence with formal convergence guarantees
- Achieves substantially faster convergence and lower communication overhead than state-of-the-art methods
- Maintains high trajectory accuracy and global consistency on loopy, noisy multi-robot graphs
- Enables fully distributed, neighbor-only computation without centralized bottlenecks
Why it matters
Provides a scalable and reliable foundation for collaborative multi-robot SLAM systems operating in complex environments with limited communication bandwidth.
Abstract
Following the relative maturity of single-robot Simul- taneous Localization And Mapping (SLAM) techniques, works ad- dressing collaborative SLAM have started emerging lately. Driven by the need for robust and scalable multi-robot systems, the community has been targeting Distributed Pose Graph Optimiza- tion (DPGO), with current DPGO methods falling into two cat- egories: optimization-based methods providing favorable conver- gence properties at the expense of excessive communication rounds among participants, and belief-propagation methods that exhibit better scalability and faster computation, albeit risking divergence on loopy and noisy graphs. Inspired by the need for more effective DPGO techniques, this work introduces Contractive Belief Shar- ing (CBS), a two-stage message-passing algorithm that combines Maximum-A-Posteriori (MAP) optimization with belief propaga- tion with a Hellinger-distance-based damping rule. In this way, CBS ensures fast and reliable convergence while maintaining lly distributed computation and communication with neighbors only. Experiments on benchmarks show that CBS reaches convergence substantially faster and more efficient and scalable than the state- of-the-art methods while maintaining high trajectory accuracy, opening up new capabilities for collaborative SLAM.