Signals, Instruments, and Systems – W13

Robotic Sensor Systems for

Environmental Monitoring
&
Overall Course Conclusion

1
 Outline
• Motivation

• Air sensing with mobile robots

• A mission example: gas source localization

• Water sensing with mobile robots

• A mission example: bacterial layer tracking

• Overall course conclusion

2
Motivation

3
 Motivation for
Spatially Dense Air Quality Monitoring
Week 12, s. 48
Air pollution in urban areas is a global concern
• affects quality of life and health
• urban population is increasing

Air pollution is highly location- and time-dependent

• traffic chokepoints and rush hours
• urban canyons and weather
• industrial installations and activities

Air pollution monitoring today

• sparse, stationary (or movable by truck) and
expensive stations
• spatial interpolation with mesoscale physics- and
chemistry-based models
4
 Motivation for
Spatially Dense Water Quality Monitoring
Fresh water quality monitoring is a global concern
• fresh water reservoirs are under pressure
• global population is increasing
• water ecosystems are not well understood

Water quality is location- and time-dependent

• natural transport phenomena, weather LéXPLORE monitoring
• interaction lake physics and biology station, Lake Geneva
• waste treatment plants, industrial installations,
agricultural activities

Water quality monitoring today

• stationary research stations leveraging vertical
profilers
• operational boats equipped with dedicated sensing
equipment
• spatial interpolation with physics- and chemistry- CIPEL monitoring
based mesoscale models boat, Lake Geneva 5
 Environmental Monitoring
Typical solution Distributed solution - size, cost
for augmentation: - number
in environmental - networked
monitoring: - mobile
- sparse sensing
- expensive
station(s)
- field estimation
via physics- and
chemistry-based
models
- possible mobility Physical field:
- built or natural
- bounded or unbounded
Week 12, s.7 - 2D or 3D 6
 OpenSense: Air Quality Monitoring
with a Mobile Sensor Systems From Week 7
Measurement data Explanatory Variables
Citizen-, consortium-, agency-operated sensors Land-use, meteorology, traffic

Exposure information High-resolution pollution maps

Personal reccomendations, health Spatiotemporally flexible, modeling;
studies, urban planning, crowdsensing emphasis on data-driven statistical modeling methods
7
 Air Sensing
with Mobile Robots

8
 Possible Application:
Gas Source Localization

9
Gas Sensing Systems

Cost Coverage Autonomy

high high low

low (handheld)
low low
high (static)

moderate high high

10
Plumes: A Tricky Field

Courtesy by L. Marques,
simulated plume

11
 Sub-Tasks for Gas Source Localization
 Plume acquisition
Source

Wind
Robot

 Plume tracking
Source Robot
Wind

 Source declaration Robot

Source

Wind
12
 Algorithms for Plume Tracking
 Gradient‐Based Algorithms
 Time‐averaged plume model x
 + Computationally efficient, intuitive xx
 ‐ Need for several samples, slow x
Wind

 Bioinspired Algorithms
 Reactive
 Finite State Machine
x
 + Computationally efficient, easy to
implement
Wind
 ‐ Not reliable in realistic environments

 Metaheuristic Algorithms
 Approximation and optimization methods x
 + Robust controller
Wind
 ‐ Learning process
13
 Algorithms for Plume Tracking
 Formation‐Based Algorithms
 Natively designed for multi‐robot systems
 Reactive
 + Computationally efficient x
 ‐ Requires inter‐robot localization
Wind
 ‐ Possibly fragile in realistic environments

 Probabilistic Algorithms
 Probability distribution of the source location
 + Flexible to the type of sensing system
 + Rich set of information as output
 + Level of uncertainty x
 ‐ Require a plume model
Wind
 ‐ Computationally expensive

14
 3D Gas Source Localization

Crazyflie V2.1
• 10 cm diameter (Nano Aerial
Vehicle)
• Flightime ~7 min
• Payload ~ 15 g
• Endowed with chemical sensing

Robot localization
• Ultra Wide Band (UWB)
• Motion Capture System (MCS)
15
 Bioinspired Algorithm
Plume finding:
Levy Walk
Plume tracking

Source declaration Plume

Plume lost
following:
Upwind surge

Success Plume
40x40x30 cm acquisition:
Plume reacquired
around inlet Spiraling

Two 3D implementations:
• [Rahbar et al., IROS 2017] on the traversing system of the wind tunnel
• [Ercolani and Martinoli, IROS 2020] on a Nano Aerial Vehicle
16
 Experiments in the Wind Tunnel

[Ercolani and Martinoli, IROS 2020] 17

Trajectories in the Wind Tunnel
UWB localization

Gas Concentration
MCS localization

Gas Concentration
18
 Performance Evaluation

Sensor Placement

Setup Wind Speed Release Rate

A 0.7 m/s High
B 0.2 m/s High
C 0.7 m/s Low
D 0.2 m/s Low

[Ercolani and Martinoli, IROS 2020] 19

Water Sensing with
Mobile Robots

20
 Underwater Environmental Sensing

 Moored / static sensors  Vertical profilers  Robotic sensor systems

 Good temporal resolution  Good vertical resolution  Good spatial resolution,
 Limited spatial resolution  Limited horizontal resolution especially in horizontal
direction
2121
 Underwater Robotics Challenges
Technical challenges
• Unavailability of radio communication

• Scarcity of external positioning references

Practical challenges
• Visibility / retrieval in case of failure
• Commercially available hardware peripherals
incompatible with small robots

2222
 A Mission Example
• SNSF Sinergia project 2015-2019, with J. Wüest and B. Ibelings
• Measuring within a thin stratified bacterial layer in Lake Cadagno (TI)
• High resolution temperature measurements within the layer to capture
spatial pattern of bacterial activity vs. local fluid dynamic conditions
• Added value of an underwater robot with respect to traditional
instruments (vertical profilers): assessment of horizontal variations of
the bacterial layer

Thin bacterial layer with

varying depth and thickness

23
The Vertex AUV and its Sensing Payload

Chlorophyll
Turbidity

Fast Temperature
[Schill et al. DARS 2016]

• Small (70 cm, 7 kg), agile vehicle

• Equipped with a suite of sensors for turbidity, chlorophyll and
high-resolution fast (20 μK, 400 Hz) temperature measurements 24
 Fully Pre-Planned Mission

25X. Trajectory estimated on-board AUV during mission in Lake Cadagno.

[Quraishi et al., ICRA 2018]

2525
 Adaptive Sampling Missions

• Turbidity measurements to detect the layer

and adapt the trajectory in real-time
• Motor vibrations can degrade the quality of the
sensitive high-resolution fast temperature sensor
measurements
• Turn off propellers for a short duration (20s)
and allow AUV to drift, to avoid motor-induced
vibrations

2626
 Adaptive Sampling Missions
• AUV adaptively locates and traces the layer
• Pink segments (grey boxes in the plot): motors turned off to
avoid motor-induced vibrations

25X. Trajectory estimated on-board AUV during mission in

Lake Cadagno.

[Quraishi et al., ICRA 2018] 2727

Conclusion

28
 Take Home Messages
• Robotic sensor systems offer controlled mobility and larger
coverage capabilities at the price of an increased complexity and
cost
• Air and water allow for the mobility to easily happen and
therefore robotic sensor systems are mainly found there
• The sensing payload depends on the environmental mission and
is typically constituted by off-the-shelf, robust sensors,
sometimes already used in canonical instruments.
• Sensing challenges are a function of the environmental mission
itself and the operability of the technology in a given setting
• Robotic sensor systems allow for closing the loop locally and
are characterized by a certain degree of on-board intelligence,
which in turn offers new opportunities in terms of
customization, automation, etc. 29
 Additional Literature – Week 13
Pointers
• https://www.epfl.ch/labs/disal/research/auvdistributedsensing/
• https://www.epfl.ch/labs/disal/research/odorsourcelocalization/
• https://www.epfl.ch/labs/disal/research/3dodorsensing/
• https://www.hydromea.com/
Papers
• Quraishi A. and Martinoli A., “Online Kinematic and Dynamic Parameter Estimation for Autonomous
Surface and Underwater Vehicles”, Proc. of the IEEE/RSJ Int. Conf. on Intelligent Robots and Systems,
September-October 2021, Prague, Czech Republic.
• Quraishi A., Bahr A., Schill F., and Martinoli A., “A Flexible Navigation Support System for a Team of
Underwater Robots”, Proc. of the Int. Symp. on Multi-Robot and Multi-Agent Systems, August 2019,
New Brunswick, USA, pp. 70-75.
• Rahbar F. and Martinoli A., “A Distributed Source Term Estimation Algorithm for Multi-Robot
Systems”, Proc. of the IEEE Int. Conf. on Robotics and Automation, May-August 2020, Paris, France,
online organization, 5604-5610.
• Ercolani C., Tang L., Humne A. A., and Martinoli A., “Clustering and Informative Path Planning for
3D Gas Distribution Mapping: Algorithms and Performance Evaluation”, IEEE Robotics and
Automation Letters, 7(2): 5310-5317, 2022. 30
Course Take Home
Messages

31
 (Intelligent) Instruments as
Specialized Embedded Systems

Vertex
e-puck

DISAL Arduino Xbee

Handheld Airborne
Sensorscope station Mapping System 32
 What These Systems Share at
their Core?

Sensing Processing

Processing Communication Visualization

Mobility Storing

In-situ instrument Transportation channel Base station

The goal of this course is to shed light on this process and blocks!
33
 What This Course Is About
• Introduction to signal processing:
– Signals, convolution, sampling and reconstruction
– Time/frequency domains, transforms, responses
– Filters
• Introduction to embedded systems:
– Sensing and sensor systems
– Communication channels
– Programming agility consolidation (interpreted vs. compiled
languages) and programming for embedded systems
• Introduction to mobile robotics:
– Basic control techniques
– Localization and sensor fusion
34
 Our Main Objectives for This Course
This course should allow you:
• To become a power user of the field instruments in environmental
engineering used nowadays (sensor networks, meteorological
stations, data loggers, etc.) and in even more so in the future
(exploratory and cleaning robots, robotic sensor systems, etc.)
• To transport your domain knowledge into code to be deployed into
programmable instruments
• To collaborate more efficiently with other engineers (e.g., computer,
electrical, mechanical, microtechnical)
• To cumulate additional background to attend, if any interest, courses
of the specialization on Environmental Sensing and Computation
including our course on Distributed Intelligent Systems
35
Thank you for your
attention!

36