Envirobot by Breadfruit

ENVIROBOT: A robot piloted by biological sensors to measure and locate pollutants in aquatic systems

ENVIROBOT is a collaborative effort between research groups from the University of Lausanne, HES‐SO, EPFL and Eawag.

An illustration: Lake Geneva • Global pollution monitored by CIPEL • Total pollutant concentration is low • Increase of specific organic pollutants – Biocide: Metalaxyl, Drug: Carisoprodol • Different hypothesis on the cause but need further measurements • Need for a smart tool to sample, measure and locate the source of pollutants in waterbody •

http://www.pugetsoundkeeper.org/

Envirobot could answer this question! Ortelli et al., 2012

In the following presentation we will show the main concept and first results of the ENVIROBOT project. We will describe in detail the different types of sensors we use, the system’s engineering and integration, and navigation of the robot.

WHAT IS THE MAIN IDEA OF THE ENVIROBOT PROJECT?

The main objective of the ENVIROBOT project is to give an existing anguilliform robot the capacity to measure pollutants in the water and to find sources of such pollution. In order to achieve this goal, the project will develop sets of miniaturized physical, chemical and biological sensors that will be integrated in the robot’s individual modules.

A variety of standard sensors for water quality characterization will be included in ENVIROBOT, such as pH, oxygen levels, turbidity and salinity. In addition, the project aims to implement chemical and biological sensors for specific types of pollutants and for “general toxicity”, as will be explained in more detail.

SENSORS

Here is an example of a water turbidity sensor developed by HES‐ SO. It is based on detection of light scattering from an emitted light beam on the bottom of a robot module, by particles in the water. Calibration series ISO7027

Here is an example of a conductivity sensor developed by the EPFL‐LEPA lab and HES‐SO. It can be used to measure dissolved metals in the water, such as lead or copper, and is mounted in the bottom of a robot module. Pb2+ in Lake Water matrix

Example: measuring atrazine with the microfluidic module Flow diagram

Data: Milica Jovic, LEPA

This type of flow‐through structures, developed by the EPFL‐LEPA lab, will be used for immunoassays or assays with bacterial reporters, and permit electrochemical readout. In the immunoassay, magnetic beads coated with antibodies capture the target compound or an artificial “competitor”, consisting of the target molecule covalently bound to horse radish peroxidase (HRP). Antibody‐target‐ HRP bound to the magnetic beads is detected by addition of an electrochemically active substrate.

A large part of the ENVIROBOT project is devoted to integrating biological sensors within the robot modules. What are biological sensors? The project will use living cells and small living water organisms. They react to compounds in the water and can signal any distress from pollution or toxicity. This gives an electrical impulse to the robot, or can be measured and interpreted quantitatively.

BIOLOGICAL SENSORS

The project uses bacteria, oocytes or fish gill cells, as well as living Daphnia as bioreporters in the robot. They can be kept inside micro‐ fluidic cages within the modules, and be exposed to the water as the robot moves. Distress signals are detected.

Daphnia as bioreporter • Setup development for Daphnia heart beat & movement observation • Microfluidic cage & perfusion • Image recording & processing • Hypothesis: Perturbation in the perfusion media with pollutants will change their heart beat or global agitation

Here is a close‐up of the Daphnia heart. Automated image analysis can be transformed into heart‐beat frequencies.

Cage development: EPFL‐LIMS4 lab. Integration into robot by HES‐SO.

Effect of ethanol added in the Daphia medium on their global agitation

Daphnia medium

Ethanol 1.4%

Daphnids agitation 27.10.2014 M2_EtOH - daphnids: 6 - threshold: 10

600

800

Daphnids agitation 27.10.2014 M4_Caffeine1mM - daphnids: 6 - threshold: 10

Legend:

600 Agitation [total passages]

200

400

500 400 300 200

Daphnia Medium EtOH 1.4% Caffeine 1mM Caffeine 10mM

100 0

Agitation [total passages]

Legend:

Daphnia Medium EtOH 1.4% Caffeine 1mM Caffeine 10mM

Time [min]

• Every two minutes the agitation is measured • When we put pollutant the daphia agitation strongly decreases

15 Time [min]

Bacteria can respond to chemicals by changing the direction of their swimming (chemotaxis). Mean individual trajectories can be recorded and transmitted to the robot’s CPU to infer sudden changes in chemical parameters in the water. Enrichment of cells to a gradient of serine

Encihment

µfluidic chip Cell enrichment at 100 x

The project will also use bacterial bioreporters. These are strains specifically designed to produce fluorescence or electrochemical response to a single chemical target (group).

50 µg/l arsenite

Day 4 Day 1

T1h50

Here an example of an arsenic‐ specific reporter strain, measured by fluorescence. The cells can be grown for long periods in microfluidic cages and exposed to a sample. PDMS µfluidic chip

Xenopus laevis oocyte based biosensor The specific interaction of pollutants with ligand‐gated ion channels expressed by oocytes triggers changes in their electrophysiological state

Upon binding a ligand, the channel opens, leading to an influx of ions. Ion influx can be measured by electrophysiology on single oocytes, here shown in a microfluidic setup. Concept and realization: UNIL‐CIG and EPFL‐ LIMS4 labs.

Single Xenopus oocyte in a cavity.

Xenopus laevis oocyte based biosensor •

Development of a non‐invasive electrophysiological measurements setup allowing the robot integration • Impedance spectroscopy based measurements

Concept and realization: UNIL‐CIG and EPFL‐ LIMS4 labs.

Finally, the project will use rainbow trout gill cells seeded and grown on the surface of biochips with embedded electrodes as bioreporters for a wide range of toxicants in the water.

Impedance (%)

Response to pentachlorophenol (in µg/L)

The response of the cells to pollutants can be followed by impedance measurements. Rainbow trout gill cells have the advantage of being very robust and active at normal water temperatures (15‐20°C).

Time [h] Concept and realization: Eawag lab.

• 37 Priority chemicals being tested • 9 wastewater treatment samples • Exposed for 24 hours, gut cell impedance loss

Lake Zurich

Tap water

L15ex

Sample 1

Sample 2

Sample 3

Sample 4

Sample 5

Sample 6

Sample 7

Sample 8

Sample 9

Time

Data: Vivian Lu, Eawag

Benchmarking fish cell lines

The development and integration of the various sensors into the robot modules is a big challenge. In addition, the modules have to communicate with the central CPU of the system and the robot has to be able to make specific navigation decisions, based on the incoming data.

SYSTEM INTEGRATION

Basis for the ENVIROBOT is the Amphibot III, a modular, portable swimming robot developed at the EPFL‐BIOROB Lab.

Swimming as an anguilliform at the water surface provides high levels of locomotion efficiency. Current autonomy is 1 h at 0.5 m/s and a range of 1.8 km. Individual modules can be easily assembled and specifically engineered for their subtasks. Each module can have a motor and electrical connections.

Individual robot modules equipped with sensors can first be tested by hooking them to a remote controlled boat, as shown here for the development and testing of a turbidity sensor made by HES‐SO.

A specific water sampling system has been built, that can filter particles above 1 µm, and pumps the filtered sample into the modules. Concept and realization: HES‐SO

Before and after filter

Passive filtering through diffusion

Transmitter

A new light weight data transmitter has been constructed, that allows communication to and from the robot over a distance of 15 m.

Receiver Concept and realization: HES‐SO

How will ENVIROBOT swim, measure and navigate? The project envisions two different modes, depending on the response time of the sensors and the deployment of the robot.

In surveying mode, the robot would follow a predefined path and can take samples at user‐defined positions. The robot can stall idle during the waiting time needed to perform the measurement, and record its data and position. For this it is equipped with GPS.

In order to find a source, the robot needs guidance, navigation and control. For this, the robot needs to interpret the data and react in an adaptive way. • Guidance decisions (“where should the robot go?”) are based on an estimation of the local trends of the concentration profile, obtained from sampled information using Bayesian Inference (Gaussian regression), • Navigation (“where is the robot?”) is performed using a Luenberger‐form observer, which low‐pass filters measures from navigation instruments, • Control (“how to pilot the robot?”) uses a predictor‐based nonlinearity, designed to account for uncertainty in the robot’s model (stemming from its hydrodynamics). Concept and realization: EPFL‐BIOROB lab.

Modeled source flow from the wastewater treatment plant in Ouchy, and the adaptive robot guidance.

Questions? Contact one of the ENVIROBOT collaborators at the posters or the demonstration desk.

THANK YOU FOR YOUR ATTENTION 30