Swarm Intelligence for Morphogenetic Engineering

TL;DR

This paper introduces a mathematical framework and programming language for artificial morphogenesis inspired by biological embryogenesis, demonstrating applications in robotic structure assembly and swarm control.

Contribution

It develops a PDE-based notation and prototype language for morphogenetic algorithms, enabling precise description and simulation of artificial morphogenesis processes.

Findings

Demonstrated morphogenetic algorithms for fiber routing and segmentation in robotics.

Showed how PDE-based specifications can derive individual control from global swarm models.

Applied smoothed particle hydrodynamics to swarm robotic control from PDEs.

Abstract

We argue that embryological morphogenesis provides a model of how massive swarms of microscopic agents can be coordinated to assemble complex, multiscale hierarchical structures. This is accomplished by understanding natural morphogenetic processes in mathematical terms, abstracting from the biological specifics, and implementing these mathematical principles in artificial systems. We have developed a notation based on partial differential equations for artificial morphogenesis and have designed a prototype morphogenetic programming language, which permits precise description of morphogenetic algorithms and their automatic translation to simulation software. Morphogenetic programming is illustrated by two examples: (1) use of a modified flocking algorithm to route dense fiber bundles between regions of an artificial cortex while avoiding other bundles; (2) use of the clock-and-wavefront model of spinal segmentation for the assembly of the segmented spine of an insect-like robot body and for assembling segmented legs on the robot's spine. Finally, we show how a variation of smoothed particle hydrodynamics (SPH) swarm robotic control can be applied to the global-to-local compilation problem, that is, the derivation of individual agent control from global PDE specifications.