JUST IN: A team of researchers presented a 3D relative localization framework for multi robot systems that combines internal angle and self displacement measurements with linear techniques, optimization on manifolds, Bayesian inference, and neural density estimation.

The result points to more accurate and robust localization, even under noise, with validation in simulations and real world drone tests.
The work proposes a distributed linear relative localization theory that allows for the estimation of positions and orientations between robots using only angles and self displacement.