IEEE Robotics & Automation Magazine - March 2012 - 71

Depth (m)

drawing conclusions from the interpolated plots. However, it has to be
−11
noted that the reference lines were
−12
recorded in a top to bottom manner,
−13
and therefore the time difference
−14
between the interpolation data 1)
−15
and the reference 2) is highest at the
−16
bottom (approximately 1.3 h). To
20 30 40 50 60 70 80 90 100 110 120 −40−20 0 20 40
compare the error of the interpolaRelative Error (%)
Distance (m)
tion to the temporal variability of
(a)
(b)
the field, we computed two rms values. The first one reflects the differFigure 10. (a) shows the data used for interpolation (black) and measurements on
ences of the interpolated values to horizontal lines for validation (blue). (b) shows the error of the interpolated values with
the reference points (10.23%). The respect to the range of the entire data set.
second one compares the raw data
points used for interpolation (black) to the closest refer- well as increasing the spatial coverage on both vertical
ence point (13.21%). The comparison of these two values and horizontal axes. Our project can be seen as a smallshows that the error of the interpolation lies within the scale sensor network that combines temporal and spatial
temporal variability of the field.
resolutions, which is often more convenient and cost efficient for lacustrine sampling. Moreover, it is worth menConclusion
tioning that our system is not restricted to lake sampling
This article presents the design of a surface vessel for the because the control functions as long as wind speed is
autonomous monitoring of water bodies and its validation lower than 15 m/s.
over a year of field testing. Three main design aspects are
Our first measurements in Lake Zurich support the
presented in detail to show that the platform has been assumptions that this water body displays a strong horispecifically tailored to suit the desired applications. First, the zontal heterogeneity in its physicochemical parameters
platform has been manufactured with lightweight materials and that these parameters can change relatively fast. These
resulting in a total weight of 120 kg, which includes 40 kg of observations raised the need for further investigations on
batteries. This allows the platform to be deployed in remote the dynamics of Planktothrix populations. For these realocations without the need for specialized infrastructure.
sons, we are routinely deploying the ASV to complement
Second, special attention has been drawn to minimize the ongoing sampling campaign of Lake Zurich that is
the probe drag while being pulled below the boat. When performed with traditional limnological techniques. We
using the depth measurement provided by the probe in also plan to develop an adaptive sampling framework in
combination with the known length of the cable, we can which measurements are analyzed online such that the
assure a reasonable localization of the probe. Even assum- layers of interest can be sampled at a higher frequency.
ing a worst-case scenario, we showed that the error on Furthermore, measurements of prior missions could be
the estimate of the probe displacement in the horizontal used to target specific sampling areas. Another sampling
direction is relatively small compared to the total travel scheme would be to detect patches of Planktothrix along a
distance of the actual sampling missions. We intend to transect across the lake and then to explore their spatial
develop in the near future a better model to reduce the extent in three-dimensional space.
resulting error.
Finally, to achieve smooth waypoint navigation, we Acknowledgments
implemented a simple, still robust control algorithm that This work is funded by the Swiss National Science Fund
ensures an accurate and repeatable line-following motion (CR22I2-130023). The authors thank Manuel Baumann
with lateral errors being smaller than 2 m, cross-validated and Oliver Baur for their construction efforts and Eugen
using a very precise external positioning system. The Loher for his constant technical support.
waypoint navigation is the core of the mission planning
system of the platform, which is straightforward enough References
so that biologists can easily operate it. The mission plan- [1] World Water Assessment Programme, "The United Nations world
ning can be adapted to the specific needs of the bio- water development report 3: Water in a changing world", UNESCO,
logists and was successfully used to conceive efficient Earthscan, Paris, London, Tech. Rep., 2009.
[2] J. Heisler, P. Glibert, D. Burkholder, J. Anderson, W. Cochlan, W.
sampling campaigns.
All these design considerations gave rise to a system Dennison, Q. Dortch, C. Gobler, C. Heil, E. Humphries, A. Lewitus, R.
capable of obtaining reliable long-term localization for Magnien, H. Marshall, K. Sellner, D. Stockwell, D. Stoecker, and M. Sudrepeated sampling at the exact same position to observe dleson, "Eutrophication and harmful algal blooms: A scientific conthe temporal evolution of a given aquatic ecosystem as sensus," Harmful Algae, vol. 8, no. 1, pp. 3-13, 2008.
MARCH 2012

IEEE ROBOTICS & AUTOMATION MAGAZINE

Table of Contents for the Digital Edition of IEEE Robotics & Automation Magazine - March 2012