A Robust Model For Predicting Collective Behavior in Large Robot Swarms [conference paper]

Conference

International Conference on Robotics and Automation (ICRA) - Robot Swarms in the Real World workshop - June 4, 2021

Authors

John Harwell (Ph.D. student), Angel Sylvester (Ph.D. student), Maria Gini (professor)

Abstract

We study the forward collective behavior problem: how to predict swarm behavior given a problem description and high level characteristics of the robot control algorithm. We present a differential equation model of swarm behavior which does not require posthoc parameter tuning or knowledge of the nature of the problem the swarm is working on to produce accurate predictions. Instead, our model computes its internal parameters directly from the problem description and robot control algorithm characteristics. We present results showing that our model accurately predicts behavior and performance in swarms of up to 12,000 robots across a wide range of object gathering scenarios

Link to full paper

A Robust Model For Predicting Collective Behavior in Large Robot Swarms

Keywords

swarm robotics