Convergence of Even Simpler Robots without Location Information

TL;DR

This paper develops and analyzes algorithms for extremely simple, oblivious robots that cannot measure distances, focusing on convergence protocols under minimal sensing assumptions, advancing the understanding of robot coordination with limited capabilities.

Contribution

The paper introduces the first deterministic self-stabilizing convergence algorithms for monoculus robots with minimal sensing, specifically under Locality Detection and Orthogonal Line Agreement models.

Findings

Algorithms guarantee convergence despite limited sensing.

Minimal assumptions are necessary; further relaxation leads to impossibility.

Complexity analysis of the proposed algorithms.

Abstract

The design of distributed gathering and convergence algorithms for tiny robots has recently received much attention. In particular, it has been shown that convergence problems can even be solved for very weak, \emph{oblivious} robots: robots which cannot maintain state from one round to the next. The oblivious robot model is hence attractive from a self-stabilization perspective, where state is subject to adversarial manipulation. However, to the best of our knowledge, all existing robot convergence protocols rely on the assumption that robots, despite being "weak", can measure distances. We in this paper initiate the study of convergence protocols for even simpler robots, called \emph{monoculus robots}: robots which cannot measure distances. In particular, we introduce two natural models which relax the assumptions on the robots' cognitive capabilities: (1) a Locality Detection ($\mathcal{LD}$) model in which a robot can only detect whether another robot is closer than a given constant distance or not, (2) an Orthogonal Line Agreement ($\mathcal{OLA}$) model in which robots only agree on a pair of orthogonal lines (say North-South and West-East, but without knowing which is which). The problem turns out to be non-trivial, and simple median and angle bisection strategies can easily increase the distances among robots (e.g., the area of the enclosing convex hull) over time. Our main contribution are deterministic self-stabilizing convergence algorithms for these two models, together with a complexity analysis. We also show that in some sense, the assumptions made in our models are minimal: by relaxing the assumptions on the \textit{monoculus robots} further, we run into impossibility results.