Approximation of Statistical Analysis and Estimation by Morphological Adaptation in a Model of Slime Mould

TL;DR

This paper demonstrates that morphological adaptation in a slime mould model can be used to perform coarse statistical analysis, estimation, and tracking, suggesting simple collective systems can extract data features through material adaptation.

Contribution

It introduces a particle-based slime mould model that uses morphological adaptation to approximate statistical properties and track moving objects.

Findings

Approximation of geometric centroid of 2D shapes

Estimation of arithmetic mean from data distributions

Dynamical tracking of moving objects with noise

Abstract

True slime mould Physarum polycephalum approximates a range of complex computations via growth and adaptation of its proto- plasmic transport network, stimulating a large body of recent research into how such a simple organism can perform such complex feats. The properties of networks constructed by slime mould are known to be in- fluenced by the local distribution of stimuli within its environment. But can the morphological adaptation of slime mould yield any information about the global statistical properties of its environment? We explore this possibility using a particle based model of slime mould. We demonstrate how morphological adaptation in blobs of virtual slime mould may be used as a simple computational mechanism that can coarsely approx- imate statistical analysis, estimation and tracking. Preliminary results include the approximation of the geometric centroid of 2D shapes, ap- proximation of arithmetic mean from spatially represented sorted and unsorted data distributions, and the estimation and dynamical tracking of moving object position in the presence of noise contaminated input stimuli. The results suggest that it is possible to utilise collectives of very simple components with limited individual computational ability (for ex- ample swarms of simple robotic devices) to extract statistical features from complex datasets by means of material adaptation and sensorial fusion.