Sparse Pose Graph Optimization in Cycle Space

The state-of-the-art modern pose-graph optimization (PGO) systems are vertex based. In this context the number of variables might be high, albeit cycles in graph (loop closures) is relatively low. For sparse problems particularly, cycle space has a significantly smaller dimension than vertices. By exploiting observation, paper we propose an alternative solution to PGO, that directly exploits space. We characterize topology as matrix, and re-parameterize problem using relative poses, which further constrained by basis graph. show minimum basis, cycle-based approach superior convergence properties against its vertex-based counterpart, terms speed global minimum. graphs, our also more time efficient vertex-based. As additional contribution work present effective algorithm compute basis. Albeit known computer science, believe not familiar robotics community. All claims validated experiments on both standard benchmarks simulated datasets. To foster reproduction results, provide complete open-source C++ implementation (Code: \url{https://bitbucket.org/FangBai/cycleBasedPGO) approach.