Large-Scale Multi-robot Collaborative Localization with Relative Observations

Hu Mingnan,Chen Bo,Yu Li
DOI: https://doi.org/10.1007/978-981-19-3998-3_13
2022-01-01
Abstract:This paper is concerned with the problem of multi-robot cooperative localization in the presence of insufficient information from proprioceptive sensors. Using only relative observations, a distributed nonlinear weighted least squares algorithm is proposed to co-localize the position and orientation of each robot. Firstly, Gauss-Newton method is utilized to decompose the nonlinear problem into a series of linear weighted least squares (WLS) steps. Then, in order to handle the linear WLS problem, an algebraic distributed method which performs finite step convergence under acyclic communication graphs is introduced, and the performance of this method is equivalent to the centralized WLS. In particular, the convergence rate of this method is only related to the diameter of the communication graph, and the computational complexity of each robot does not increase as the communication graph becomes large. Moreover, it turns out that the proposed collaborative localization algorithm is exponentially convergence under cyclic communication graphs. Finally, numerical experiments are used to verify the effectiveness and superiority of the proposed method.
What problem does this paper attempt to address?