Date of Award
2026
Degree Type
Thesis
Degree Name
Master of Science in Mechanical Engineering and Applied Mechanics
Department
Mechanical, Industrial and Systems Engineering
First Advisor
Yang Lin
Abstract
Passive wake signatures in fluid flows can support perception and low-rate communication in swarms. In cluttered or contested underwater environments, conventional acoustic, radio, and optical links can be power-hungry, intermittent, or undesirable when low observability is required. Wake-mediated cues offer a local, directional channel that does not require line of sight because each agent naturally sheds coherent vortices that persist downstream and can be sampled by followers with only a small number of probes.
This study evaluates whether sparse downstream probes are sufficient to infer agent attributes and decode simple messages from wakes generated by established source shapes. Computational fluid dynamics simulations are conducted in COMSOL Multiphysics for incompressible flows at moderate Reynolds number (Re = 200) to generate wake datasets from circular and elliptical bodies arranged as three-body platoons in linear and V-formations. Probe time series of velocity magnitude are processed through a supervised learning pipeline in which a Random Forest classifier estimates shape and an XGBoost regressor estimates geometric scale from compact signal features. For circular sources with diameters from 1.0 to 3.0 m and probe distances from 1 to 5 m behind the leader, shape-classification accuracy remains near 0.93, while size-estimation error decreases from roughly 24 to 28 percent at 1 m to roughly 9 to 19 percent at 5 m. These trends suggest that farther-wake measurements provide a more stable regression signature because near-wake probes are more strongly affected by shear-layer roll-up, proximity effects, and follower-body interference.
In addition to inference, the wake-mediated messaging concept in which a leader fin is toggled to introduce a transient perturbation into the wake is demonstrated in this study. A two-stage decoder based on persistence and cross-probe consistency limits false positives and supports a small message alphabet in simulation. To isolate the main mechanisms and keep the dataset tractable, the study focuses on two-dimensional, noise-free simulations at moderate Reynolds number, while also discussing how the conclusions are expected to change in higher-Reynolds- number, three-dimensional, and noisy settings. The results indicate that the fluid itself can serve as a medium for decentralized sensing and low-observability signaling in close-range swarms.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Varela, Bryan, "PASSIVE COMMUNICATION ACROSS DIVERSE SWARM FORMATIONS AND SCALES UTILIZING WAKE SIGNATURES FOR MESSAGING AND OBJECT INFERENCE" (2026). Open Access Master's Theses. Paper 2713.
https://digitalcommons.uri.edu/theses/2713