Applications to Biological Networks of Adaptive Hagen-Poiseuille Flow on Graphs

TL;DR

This paper investigates a physically consistent model of adaptive Hagen-Poiseuille flow on graphs inspired by Physarum polycephalum, demonstrating its ability to produce biologically realistic and efficient networks, with applications to real-world transport systems.

Contribution

It introduces and analyzes a novel adaptive flow model on graphs that replicates key features of Physarum networks and optimizes for efficiency and cost.

Findings

Model networks follow Murray's law at steady state.

Networks exhibit Physarum-like structures, including peristalsis and shuttle streaming.

Algorithms outperform real transport networks in efficiency and some cost metrics.

Abstract

Physarum polycephalum is a single-celled, multi-nucleated slime mold whose body constitutes a network of veins. As it explores its environment, it adapts and optimizes its network to external stimuli. It has been shown to exhibit complex behavior, like solving mazes, finding the shortest path, and creating cost-efficient and robust networks. Several models have been developed to attempt to mimic its network's adaptation in order to try to understand the mechanisms behind its behavior as well as to be able to create efficient networks. This thesis aims to study a recently developed, physically-consistent model based on adaptive Hagen-Poiseuille flows on graphs, determining the properties of the trees it creates and probing them to understand if they are realistic and consistent with experiment. It also intends to use said model to produce short and efficient networks, applying it to a real-life transport network example. We have found that the model is able to create networks that are consistent with biological networks: they follow Murray's law at steady state, exhibit structures similar to Physarum's networks, and even present peristalsis (oscillations of the vein radii) and shuttle streaming (the back-and-forth movement of cytoplasm inside Physarum's veins) in some parts of the networks. We have also used the model paired with different stochastic algorithms to produce efficient, short, and cost-efficient networks; when compared to a real transport network, mainland Portugal's railway system, all algorithms proved to be more efficient and some proved to be more cost-efficient.