Bottom-up robust modeling for the foraging behavior of Physarum polycephalum

TL;DR

This paper develops multi-agent models to understand the minimal rules behind Physarum polycephalum's robust network formation, revealing insights into self-organization and potential bio-inspired decentralized network strategies.

Contribution

It introduces a hierarchy of models that analyze the minimal mechanisms enabling Physarum's foraging and network formation behaviors.

Findings

Models show robust pattern formation aligned with experimental data

Sensitivity analysis reveals parameter influences on network robustness

Hierarchical models facilitate understanding of minimal self-organizing rules

Abstract

The true slime mold \textit{Physarum polycephalum} has the remarkable capability to perform self-organized activities such as network formation among food sources. Despite well reproducing the emergence of slime networks, existing models are limited in the investigation of the minimal mechanisms, at the microscopic scale, that ensure robust problem-solving capabilities at the macroscopic scale. To this end, we develop three progressively more complex multi-agent models to provide a flexible framework to understand the self-organized foraging and network formation behaviors of \textit{Physarum}. The hierarchy of models allows for a stepwise investigation of the minimal set of rules that allow bio-inspired computing agents to achieve the desired behaviors on nutrient-poor substrates. By introducing a quantitative measure of connectedness among food sources, we assess the sensitivity of the model to user-defined and bio-inspired parameters, as well as the robustness of the model to parameter heterogeneity across agents. We ultimately observe the robust emergence of pattern formation, in line with experimental evidence. Overall, our study sheds light onto the basic mechanisms of self-organization and paves the way towards the development of decentralized strategies for network formation in engineered systems, focusing on trade-offs between biological fidelity and computational efficiency.