Insect-inspired Visually-guided Decentralized Swarming

TL;DR

This paper introduces an insect-inspired, vision-based decentralized swarming method that enables multi-agent coordination using optic flow sensing, without relying on explicit communication or position data, with proven convergence guarantees.

Contribution

It develops a novel optic flow sensing framework combined with distributed feedback for vision-guided swarming, extending Cucker-Smale flocking models with rigorous convergence analysis.

Findings

Achieved asymptotic convergence in vision-guided swarming

Developed a multi-agent optic flow sensing framework

Extended flocking models to vision-based control

Abstract

This paper addresses the need for fast, lightweight, vision-guided swarming under limited computation and no explicit communication network or position source. The study develops a multi-agent optic flow sensing framework, then integrates perfect information distributed feedback with optic flow sensing to create an analogous visually-guided feedback path for idealized inter-agent velocity and distance structures. The Cucker-Smale flocking example is used to develop vision-guided swarming with rigorous asymptotic convergence guarantees, including under ignorance of agent size.