Distributed network of smartphone sensors: a new tool for scientific field measurements

TL;DR

This work addresses the challenge of obtaining multi-point, time-synchronized measurements in field settings by leveraging a fleet of identical smartphones as autonomous, time-synchronized MEMS-based sensors. The authors develop Gobannos for per-device acquisition and clock stabilization, and PhoneFleet to orchestrate remote control and data collection, achieving a clock synchronization precision of about $60~\mu\mathrm{s}$. They validate the approach through two demonstrations: a pendulum-chain in turbulent wind revealing dual dispersion branches and a wave-buoy network on ice illustrating wave attenuation over tens of meters, with an attenuation length of $\ell_c = 37.3$ m and a principal wave frequency near $f_0 = 0.3 \mathrm{Hz}$. The results show that large smartphone fleets can provide high-temporal-resolution, spatially distributed measurements at a fraction of the cost of traditional sensor networks, enabling accessible large-scale experimental studies in environmental and geophysical contexts.

Abstract

Smartphones sensors are now commonly used by a worldwide audience thanks to their availability, high connectivity, and versatility. Here, we present a methodology to use a collection of smartphones, namely a fleet, as a distributed network of time-synchronized mechanical sensors. We first present the mechanical tests we develop to evaluate the smartphone sensor accuracy. We then describe how to use efficiently a distributed network of smartphones as autonomous sensors. We use a combination of an Android application hosted on each phone (Gobannos), and a server application (Phonefleet) on a controlling host to perform the tasks in parallel remotely. We implement in particular a time synchronization protocol based on UDP communication. We achieved an accuracy of the smartphone clock synchronisation of 60 microseconds. Using two test cases in realistic outdoor conditions, we eventually prove the reliability of a smartphone fleet to measure mechanical wave measurements in field conditions.

Figures (12)

• Figure 1: (a) Sketch of the smartphone coordinate system ($0,x,y,z$) with the origin located on the top-right up corner of the phone's bounding box. The accelerometer sensor is located at point $S$. (b) Sketch of the turning-table setup used for the sensor tests.
• Figure 2: Noise spectrum of accelerometer (a), gyroscope (b) and magnetometer (c) computed on a noise recordings of 5 minutes (red curve). Each black line corresponds to a single phone recording. Note that the sampling frequency is around 400 Hz for the accelerometer, while it is only 50 Hz for both gyroscope and magnetometer.
• Figure 3: Determination of the accelerometer sensor location along the three main smartphone axis. a), b) and c) Measurements of respectively $a_x$, $a_y$, $a_z$ as a function of the sliding position $x$ (resp. $y$, $z$), for a rotation axis aligned with $y$ (resp. $x$ and $x$). In the three cases we observe a linear relationship, which intersects zero at the sensor position $x_S$ (resp. $y_S$, $z_S$) denoted by the red stars.
• Figure 4: Conceptual sketch for (a) fleet remote control installation protocol and (b) fleet field deployment protocol.
• Figure 5: a) Image of the wind tunnel measurement setup. The chain of smartphones is placed in the symmetry plane of the measuring section, in suspension with metal trestles. b) Technical drawing of the rod-mount suspension system and its assembly. $\ell_S=50.6$ mm and $\ell_G=111.8$ mm indicate the rod to the accelerometer distance and the rod to the center of gravity distance. c) Spatiotemporal chart of the tangential acceleration $a_z$ of the chain of telephones, for a wind speed $U=9.3$ m/s. Inset: temporal signal of the tangential $a_z$ (blue) and radial (green) acceleration $a_y$ of the smartphone $\#50$ located at around $x=8$ m. Both the signal duration and smartphone position are indicated by the white dashed line.