Efficient, Incremental Subgraph Selection with Budget-Awareness

This work describes a novel, rigorous method to improve the cost-efficiency of BA-based VSLAM back-end, which is essential for SLAM applications with computation limit.

An efficient algorithm is developed to select scale-reduced graph optimized in local BA with conditioning preservation. The budget of local BA, as well as the desired scale of selected graph, are determined with budget-awareness. The proposed algorithm is integrated into a state-of-the-art VSLAM system. Superior performance is achieved under a variety of computation limits, when compared against state-of-the-art VSLAM systems.

The good graph feature is open-sourced at our SLAM stack: https://github.com/ivalab/gf_orb_slam2

Figures generated using full evaluation results can be accessed at: https://github.com/ivalab/FullResults_GoodGraph