Online Dynamic SLAM with Incremental Smoothing and Mapping
Jesse Morris, Yiduo Wang, Viorela Ila
AI summary
Problem
Existing dynamic SLAM frameworks rely on computationally intensive batch optimization, making them unsuitable for real-time applications. Formulating dynamic SLAM for incremental inference is particularly challenging due to dense graph connectivity that scales with the number of visible objects.
Approach
The authors introduce a Hybrid factor-graph formulation that merges object-centric and world-centric representations to preserve sparsity, alongside a Parallel-Hybrid architecture that decouples static and dynamic estimation into independent, parallelizable graphs solved via incremental smoothing.
Key results
- Novel Hybrid factor-graph formulation combining object and world-centric representations
- First application of incremental smoothing and mapping (iSAM2) to dynamic SLAM
- Parallel-Hybrid architecture decoupling static and dynamic components for efficient parallel computation
- 5× computational speed-up over state-of-the-art baselines with equal or better accuracy
Why it matters
Enables real-time, safety-critical robotic navigation in complex dynamic environments by making dynamic SLAM computationally feasible for online deployment.
Abstract
Dynamic SLAM methods jointly estimate for the static and dynamic scene components. However, existing ap- proaches, while accurate, are computationally expensive and unsuitable for online applications. In this work, we present a novel factor-graph formulation and system architecture for Dynamic SLAM that inherently supports incremental opti- misation and online estimation. This represents the first for- mulation explicitly designed to leverage incremental inference methods in the dynamic setting. On multiple datasets, we demonstrate that our method achieves camera pose and object motion accuracy equal to or better than state-of-the-art. We further analyse the structural properties of our approach to demonstrate its scalability and provide insight regarding the challenges of solving Dynamic SLAM incrementally. Finally, we show that our formulation leads to problem structure well- suited to incremental solvers, and our system architecture further enhances performance, achieving a 5× speed-up over existing methods. Code is open-sourced at https://github.com/ ACFR-RPG/DynOSAM.