Organisms relying on chemical cues for navigation face significant
challenges due to complexity in the environment. For instance, atmospheric
turbulence dilutes and mixes odor signals with other scents and
clean air, providing only weak, intermittent cues for insects like moths
to navigate. Despite these challenges, many species develop effective
strategies to locate distant targets in complex environments. This
raises a key question: how are the sporadic chemical signals utilized
to implement efficient source-localization strategies? The searcher’s
memory of previously detected signals plays a vital role in this process.
Current algorithms typically require continuous memory spaces
with high dimensionality, which may impede optimization and complicate
interpretation.
In this research, we demonstrate through a computational modeling of
the source localization problem that finite-state controllers, simple algorithmic
devices with minimal memory requirements, are rich enough
to explain various behavioral patterns observed in nature, first in the
context of olfactory search. The controller’s memory states emerged
to encode dual information streams: temporal data functioning as a
clock, and spatial data serving as a map. In the microscale level, we
developed a finite-state controller for E. coli chemotaxis that achieves
precise adaptation and exhibits positive responses to increasing stimuli.
Lastly, we extend the olfactory search problem to analyze sourcetracking
in an alternative context: a porous medium characterized
by chaotic flow patterns, where agents must simultaneously learn to
circumvent obstacles while localizing a chemical signal source. Our
findings demonstrate that finite-state controllers are simple yet powerful
tools for understanding behavioral patterns in diverse navigation
scenarios.
