1: \begin{abstract}

2: The goal of the present paper is to highlight the fundamental differences of so-called \emph{synchronization} or \emph{consensus} algorithms when the agents to synchronize evolve on a \emph{compact homogeneous manifold} (like the circle, sphere or the group of rotation matrices), instead of a vector space. For the benefit of understanding, the discussion is restricted to the circle. First, a fundamental consensus algorithm on $\mathbb{R}^n$ is reviewed, from which continuous- and discrete-time synchronization algorithms on the circle are deduced by analogy. It is shown how they are connected to Kuramoto and Vicsek models from the literature. Their convergence properties are similar to vector spaces only for specific graphs or if the agents are all located within a semicircle. Examples are proposed to illustrate several other possible behaviors. Finally, new algorithms are proposed to recover (almost-)global synchronization properties.

3: \end{abstract}

4: