1: \begin{abstract}

2: 	We propose fast and communication-efficient distributed algorithms for

3: 	rotation averaging and translation recovery problems

4: 	that arise from multi-robot simultaneous localization and mapping (SLAM)

5: 	and distributed camera network localization applications.

6: 	Our methods are based on theoretical relations between the \emph{Hessians} of the underlying Riemannian optimization problems 

7: 	and the \emph{Laplacians} of suitably weighted graphs.

8: 	We leverage these results to design a distributed solver that performs approximate second-order optimization

9: 	by solving a \emph{Laplacian system} at each iteration.

10: 	Crucially, our algorithms permit robots to employ \emph{spectral sparsification} to sparsify intermediate dense matrices before communication, 

11: 	and hence provide a mechanism to trade off accuracy with communication efficiency with provable guarantees.

12: 	We perform rigorous theoretical analysis of our methods and prove that they enjoy (local) \emph{linear} rate of convergence on the problems of interest.

13: 	Numerical experiments show that the proposed methods converge

14: 	to high-precision solutions in a few iterations

15: 	and that they are significantly more communication-efficient compared to baseline second-order solvers.

16: \end{abstract}

17: