Learning to Swarm with Knowledge-Based Neural Ordinary Differential Equations

TL;DR

This paper introduces a framework using knowledge-based neural ordinary differential equations to learn decentralized robot controllers from swarm trajectory data without action labels, enabling scalable flocking behavior reproduction.

Contribution

It presents a novel method combining neural ODEs with known agent dynamics to learn decentralized controllers solely from state observations, without requiring action data.

Findings

Successfully learned controllers reproduce flocking behavior.

Controllers scale to larger swarms with more robots.

Efficient training leveraging swarm network structure.

Abstract

Understanding decentralized dynamics from collective behaviors in swarms is crucial for informing robot controller designs in artificial swarms and multiagent robotic systems. However, the complexity in agent-to-agent interactions and the decentralized nature of most swarms pose a significant challenge to the extraction of single-robot control laws from global behavior. In this work, we consider the important task of learning decentralized single-robot controllers based solely on the state observations of a swarm's trajectory. We present a general framework by adopting knowledge-based neural ordinary differential equations (KNODE) -- a hybrid machine learning method capable of combining artificial neural networks with known agent dynamics. Our approach distinguishes itself from most prior works in that we do not require action data for learning. We apply our framework to two different flocking swarms in 2D and 3D respectively, and demonstrate efficient training by leveraging the graphical structure of the swarms' information network. We further show that the learnt single-robot controllers can not only reproduce flocking behavior in the original swarm but also scale to swarms with more robots.