Hostname: page-component-7b7567c8c-2n9c9 Total loading time: 0 Render date: 2026-11-17T02:54:05.255Z Has data issue: false hasContentIssue false

Optimal sensor placement for artificial swimmers

Published online by Cambridge University Press:  10 December 2019

Siddhartha Verma
Affiliation:
Computational Science and Engineering Laboratory, Clausiusstrasse 33, ETH Zürich,CH-8092, SwitzerlandDepartment of Ocean and Mechanical Engineering, Florida Atlantic University, Boca Raton,FL33431, USAHarbor Branch Oceanographic Institute, Florida Atlantic University, Fort Pierce, FL34946, USA
Costas Papadimitriou
Affiliation:
Department of Mechanical Engineering, University of Thessaly, Pedion Areos, GR-38334Volos, Greece
Nora Lüthen
Affiliation:
Computational Science and Engineering Laboratory, Clausiusstrasse 33, ETH Zürich,CH-8092, Switzerland
Georgios Arampatzis
Affiliation:
Computational Science and Engineering Laboratory, Clausiusstrasse 33, ETH Zürich,CH-8092, Switzerland
Petros Koumoutsakos*
Affiliation:
Computational Science and Engineering Laboratory, Clausiusstrasse 33, ETH Zürich,CH-8092, Switzerland
*
†Email address for correspondence: petros@ethz.ch

Abstract

Natural swimmers rely for their survival on sensors that gather information from the environment and guide their actions. The spatial organization of these sensors, such as the visual fish system and lateral line, suggests evolutionary selection, but their optimality remains an open question. Here, we identify sensor configurations that enable swimmers to maximize the information gathered from their surrounding flow field. We examine two-dimensional, self-propelled and stationary swimmers that are exposed to disturbances generated by oscillating, rotating and D-shaped cylinders. We combine simulations of the Navier–Stokes equations with Bayesian experimental design to determine the optimal arrangements of shear and pressure sensors that best identify the locations of the disturbance-generating sources. We find a marked tendency for shear stress sensors to be located in the head and the tail of the swimmer, while they are absent from the midsection. In turn, we find a high density of pressure sensors in the head along with a uniform distribution along the entire body. The resulting optimal sensor arrangements resemble neuromast distributions observed in fish and provide evidence for optimality in sensor distribution for natural swimmers.

Information

Type
JFM Papers
Copyright
© 2019 Cambridge University Press
Figure 0

Figure 1.(a) The lateral line in juvenile zebrafish, with neuromasts visible as bright dots on the body surface (adapted with permission from Sapède et al. (2002)). We observe a high density of neuromasts in the head and the tail, with sparser distribution along the midsection. (b) A schematic representation of the distribution of mechanoreceptors along the fish body. (c) The neuromasts bend in response to flow, which generates a neuronal response from sensory cells located at the base (adapted with permission from Kottapalli et al. (2013)).

Figure 1

Figure 2.(a) A larva-shaped swimmer detecting disturbances generated by a rotating cylinder (angular velocity $=$ $10~\text{rotations}~\text{s}^{-1}$). Regions with positive vorticity are coloured in red, and those with negative vorticity are coloured in blue. (b) An adult-shaped swimmer detecting an oscillating cylinder (amplitude $=$ $0.075L$, frequency $=$ 10 Hz). Animations for these two cases are shown in supplementary Movie 1 and Movie 2 and are available online at https://doi.org/10.1017/jfm.2019.940.

Figure 2

Figure 3.Set-up used for determining the optimal sensor distribution on a fish-like body swimming past a cylinder located within the rectangular area, which is referred to as the prior-region. The sensor placement algorithm attempts to find the best arrangement of sensors that allows the swimmer to identify the correct cylinder position with minimal uncertainty.

Figure 3

Figure 4.Snapshots of the vorticity field around a static larva profile in the presence of a horizontally oscillating cylinder. The snapshots are taken at regular intervals over a single oscillation period, with positive vorticity shown in red and negative vorticity shown in blue. A corresponding animation is shown in supplementary movie 3.

Figure 4

Figure 5.Utility plots for a stationary, larva-shaped body with (a) oscillating and (b) rotating cylinders. The curves indicate the utility for placing the first shear stress sensor at a given location $s$. The utility curves were not computed in the region $0.95, to avoid potential numerical issues resulting from sharp corners at the tail. (c,e) Standard deviation of horizontal and vertical velocity caused by oscillating cylinders, with larger deviation shown in yellow and lower values shown in black. The standard deviation was computed across nine distinct simulations (six time-snapshots recorded in each simulation), with a single oscillating cylinder placed at nine locations uniformly in the prior-region. (d,f) Standard deviation of velocity components for the rotating cylinders. (g) Optimal sensor distribution determined using sequential placement. Sensors for detecting oscillating cylinders are shown as black squares, whereas those for detecting rotating cylinders are shown as blue circles. The numbering indicates the sequence determined by the optimal placement algorithm.

Figure 5

Figure 6.(a) Utility curves for placing the first three sensors on a static larva that detects oscillating cylinders (figure 4). The solid green curve corresponds to $\hat{U} _{1}(s)$, the dashed purple curve to $\hat{U} _{2}(s)$ and the red dash-dot curve to $\hat{U} _{3}(s)$. (b) The optimal utility $\hat{U} _{n}^{\star }$ for the $n$th sensor can be determined as $\max _{s}\hat{U} _{n}(s)$ using the curves shown in panel (a) (see also (2.9)).

Figure 6

Figure 7.Snapshots of the vorticity field around a static larva in the wake of a D-shaped cylinder with diameter $0.5L$. The D-cylinder is oriented at a $10^{\circ }$ angle with respect to a uniform horizontal flow to promote vortex shedding. The snapshots are shown at regular time intervals, with positive vorticity shown in red and negative vorticity shown in blue. A corresponding animation is shown in Movie 4.

Figure 7

Figure 8.(a) Utility curve $\hat{U} _{1}(s)$ for larvae in a D-cylinder’s wake. (b) Sequential placement of 20 sensors, with the order of placement shown.

Figure 8

Figure 9.(a) Utility curves for the first shear stress sensor, $\hat{U} _{1}(s)$, on a larva-shaped swimmer (black squares, oscillating cylinders; blue circles, rotating cylinders). (b) Sequential placement of 20 sensors along the body, with the order of placement shown. (c) Utility curves for an adult-shaped swimmer. (d) Sensor placement for the adult, with results from horizontal and rotating disturbances shown separately for clarity.

Figure 9

Table 1.Optimal distribution of the first 10 shear stress sensors for the self-propelled swimmers. The body has been divided into three distinct segments: the head ($0\leqslant s/L<0.2$), the midsection ($0.2\leqslant s/L<0.6$) and the posterior ($0.6\leqslant s/L\leqslant 1$).

Figure 10

Figure 10.(a) Utility curve for the first shear stress sensor, $\hat{U} _{1}(s)$, on a larva-shaped swimmer, using a combination of all five flow configurations described in the paper. (b) Sequential placement of 20 shear stress sensors along the body.

Figure 11

Figure 11.(a) Utility curve for the first pressure gradient sensor, $\hat{U} _{1}(s)$, on a larva-shaped swimmer, using a combination of all five flow configurations described in the paper. (b) Sequential placement of 20 pressure gradient sensors along the body.

Figure 12

Figure 12.Plots showing the probability that the signal being measured by a self-propelled swimmer originates from a particular location in the prior-region. (a,c,e) Swimmer using flow measurements from optimal sensor locations indicated on the body, panels (b,d,f) show estimates made by a swimmer using suboptimal sensor distributions. Brighter areas indicate regions of higher probability. The relevant sensor arrangement is shown using red ‘$\times$’ symbols on the swimmers’ bodies. The actual position of the signal-generating cylinder is marked with red diamonds. (a,c,e) Probability distributions computed using measurements from 1, 3 and 5 optimal sensors on both sides of the body. (b,d,f) Probability distributions for 1, 3 and 5 uniformly distributed sensors on both sides.

Verma et al. supplementary movie 1

A larva-shaped swimmer detecting disturbances generated by a rotating cylinder.

Download Verma et al. supplementary movie 1(Video)
Video 8.9 MB

Verma et al. supplementary movie 2

An adult-shaped swimmer detecting an oscillating cylinder.

Download Verma et al. supplementary movie 2(Video)
Video 8.2 MB

Verma et al. supplementary movie 3

Vorticity field around a static larva in the presence of a horizontally oscillating cylinder.

Download Verma et al. supplementary movie 3(Video)
Video 2.7 MB

Verma et al. supplementary movie 4

Vorticity field around a static larva in the wake of a D-shaped__cylinder.

Download Verma et al. supplementary movie 4(Video)
Video 5.3 MB