Guerlédan

De nouvelles inspections du barrage ont été réalisées aujourd’hui dans le cadre d’un projet d’inspection de structures sous-marines. Avec le développement des éoliennes offshores, la surveillance régulière de telles installations devient un enjeu important. L’idée de ce projet est de réaliser, avec un robot sous-marin autonome, une inspection des structures immergées en utilisant l’architecture du site pour se repérer. Les signaux GNSS ne se propageant pas sous l’eau, l’utilisation de caméras optiques et de sonars devrait permettre de reconnaitre les structures et donc de retrouver où le robot est positionné sous l’eau.

Les robots sous-marins peuvent également travailler en équipe. Nous nous sommes intéressés au concept d’un robot sous-marin compagnon qui servirait à éclairer la zone de travail du robot principal ou à l’assister quand des manipulations complexes sont nécessaires. Une des difficultés est de ne pas emmêler les câbles lorsque ces robots sont reliés à la surface par un ombilical. Il faut également que le robot compagnon soit capable de suivre correctement les mouvements et les consignes du robot principal ce qui est un problème difficile sous l’eau et avec une faible visibilité.

L’acquisition des données de profondeur pour la fabrication de cartes marines va fortement évoluer grâce à l’utilisation de robots autonome. Le problème de localisation de ces robots est donc crucial. Pour améliorer la localisation du robot, un projet très innovant s’intéresse à l’analyse des zones que le robot aura vues deux fois, ce qui est souvent le cas pour garantir que toute une zone a été couverte. Si la trajectoire estimée du robot est cohérente, les mêmes choses doivent être observées la seconde fois que le robot passe sur une zone. Dans le cas contraire, il est possible de réestimer la trajectoire du robot et donc de l’améliorer. Différentes zones du lac ont été étudiées pour obtenir plusieurs types de fond et tester les limites des algorithmes.

Des marégraphes qui mesurent la hauteur d’eau ont été installés au fond du lac à des positions stratégiques. Ils cherchent à déterminer si des seiches sont présentes dans le lac de Guerlédan. Ce phénomène correspond à l’apparition sous l’effet du vent d’une onde particulière, comme de la houle, qui dépend de la forme du lac. Des capteurs de vents ont également été installés pour comparer les modèles aux données expérimentales.

Les DDboats ont progressé en autonomie cette semaine. En février prochain, un challenge avec les étudiants d’Erasmus Mundus MIR à Guerlédan consistera en plusieurs courses de bateaux autonomes. Les roboticiens de deuxième année ont affuté leurs algorithmes pour être les meilleurs. Au programme, il faudra effectuer des parcours entre plusieurs bouées ou passer par des portes à un instant précis. Cela requiert de prendre en compte la trajectoire, la vitesse du bateau ainsi que les paramètres environnementaux tel le vent.

Pour conclure cette première semaine sur le lac de Guerlédan, nous notons que la transmission de connaissances entre deuxième et troisième année a déjà bien commencé !

Aujourd’hui, nous avons redécoré le barrage de Guerlédan avec des mires. Ces disques noirs et blancs servent de référence pour recaler les images utilisées dans les algorithmes de photogrammétrie. Nous sommes montés en haut du barrage pour installer de manière précise ces mires en mesurant leur position à l’aide d’un GPS. D’autres mires ont également été installées sur la paroi du barrage.

Un ROV a également été déployé pour acquérir des images sous-marines en plus des images aériennes obtenues depuis un semi-rigide. Nous avons pu observer à 35m de profondeur, grâce au ROV, la vanne de vidange de fond du barrage.

Un robot bateau autonome – USV, Unmanned Surface Vehicles, a navigué sur le lac. Il a pour mission de trouver un objet immergé dans une zone de faible fond à l’aide d’une caméra sous-marine positionnée sous le robot. L’innovation liée à ce projet est de garantir formellement que le robot a pu observer complètement une zone. Cela nécessite de prendre en compte les incertitudes de positionnement du bateau ainsi que la portée de détection de la caméra. Être capable de garantir qu’une zone a été complètement explorée est essentiel dans des opérations de recherche et peut être transposée à d’autres milieux que le sous-marin.

Le lac de Guerlédan est également un bon environnement de test avant des expérimentations en mer. C’est le cas du projet qui cherche à comprendre le transport des larves d’huitres. Des capteurs de turbidité et de température sont utilisés afin de construire un modèle numérique simulant les courants et la température. Le développement des larves étant sensible à la température, il faut combiner cette information à celle des courants pour construire une simulation permettant de comprendre l’évolution et la localisation des populations d’huitres. Les protocoles d’acquisition des capteurs sont donc testés à Guerlédan avant d’être réalisés en pleine mer ce qui est plus difficile.

La faune du lac n’est pas en reste, un projet s’intéresse également à caractériser les espèces et la quantité de poisson dans le lac. Les signaux de trois sondeurs monofaisceau, émetteur-récepteur acoustique mono voix, sont utilisés et traités à l’aide d’algorithmes particuliers pour détecter les poissons. Les signaux obtenus sont représentés sous forme d’échogrammes qui présentent des structures particulières en fonction des espèces de poissons. Ce projet comporte également une dimension artistique : des étudiants de l’UBO en master ingénieur du son vont utiliser ces échogrammes pour générer un morceau de musique. Leurs travaux seront diffusés lors d’un festival.

La structure physique du lac présente aussi des variations importantes de température : environ 18°C en surface et 7°C au fond. Lorsque des instruments acoustiques sont utilisés, ces variations perturbent la trajectoire des ondes sonores. C’est pourquoi, lorsque les hydrographes utilisent des sonars pour cartographier le lac, ces variations physiques du milieu doivent être prises en compte pour donner une profondeur correcte. Un des projets s’intéresse à optimiser cette phase de correction. Au lieu de mesurer précisément la structure du lac à l’aide d’un capteur que l’on fait descendre jusqu’au fond, l’idée est de réaliser la correction directement avec les mesures acoustiques des sonars. Cela nécessite des algorithmes particuliers d’optimisation à appliquer de manière innovante sur les données acquises.

Nous avons également eu la visite de l’entreprise Kopadia qui co-encadre une thèse CIFRE en robotique. Ils ont déployé un AUV pour tester des algorithmes en prévision d’une mission d’inspection.

A focus on a project : Autonomous bathymetric surveys

Six robotics and hydrography students in Guerlédan aim to map the seabed of the lake using an autonomous surface drone: we call that a bathymetric survey.

Bathymetry is the science of measuring the depths and relief of the ocean to determine the topography of the sea floor. ENSTA students have a companion named Ulysses: a semi-autonomous catamaran developed jointly by Texys Marine, a company manufacturing surface drones and sensors, and ENSTA Bretagne.

Currently, no device can carry out a bathymetric survey in complete autonomy. Instead, survey boats are brought back to the lab to check the results. In case of an inconsistency, or in case of an error, scientists have to repeat the entire mission. Our team’s main objective is to develop algorithms that allow Ulysses to carry out a mission in complete autonomy, that is to topographically map the seabed. For this purpose, Ulysses will be equipped with a multi-beam sounder – a type of sonar that is used to map the seabed -, a GPS and an inertial unit – an electronic device that measures a body’s specific force and its orientation. In addition, it must be able to perform real-time data quality control, which will indicate where it will have to come back to for additional surveys.

This project is at the crossroads between two different fields: Robotics and Hydrography. On one hand, a lot of middleware – interfaces between software and hardware – are used, such as ROS (Robot Operating System). On the other hand, in hydrography, most software are under Windows. Managing communication between these two worlds is a real ordeal for our engineers!

Focus on : Assessment and correction of systematic errors

Four third-year students have been taking measurements with their boat’s sonar since the beginning of the week. They aim at mapping the bed of the lake and to correct and estimate the errors of the sonar caused by the variations of sound celerity between the layers of water.

Roboticists in action

Meanwhile, some 2^nd year robotics students are testing their programs in real-life conditions. Many factors such as wind, waves and satellite position but also mainly coding errors will impact the trajectory of these small boat bots.

As the first week of the project draws to an end, students take stock of the difficulties linked to working in groups, depending on unstable environmental conditions, facing electronic glitches which simulation tests would not allow them to predict.

A few of them enjoyed discovering the Guerlédan lake at the best moment of the day!

Focus on a crossover project

Today we followed six robotics and hydrography students in Guerlédan. Their aim is to map the seabed of the lake using an autonomous surface drone: we call that a bathymetric survey. Bathymetry is the science of measuring the depths and relief of the ocean to determine the topography of the seafloor. ENSTA students have a companion named Ulysses: it is a semi-autonomous catamaran developed jointly by Texys Marine, a company which manufactures surface drones and sensors, and ENSTA Bretagne. Currently, no device can carry out a bathymetric survey in complete autonomy. Instead, survey boats are brought back to the lab to check the results. In case of an inconsistency, or in case of an error, scientists have to repeat the entire mission.

Our team’s main objective is to develop algorithms that will allow Ulysses to map the seabed in complete autonomy. For this purpose, Ulysses will be equipped with a multibeam sounder – a type of sonar that is used to map the seabed, – a GPS and an inertial unit – an electronic device that measures a body’s specific force and its orientation. In addition, it must be able to perform real-time data quality control, which will indicate where it will have to come back to for additional surveys. This project is at the crossroads between two different fields: Robotics and Hydrography. Robotics uses a lot of middleware – interfaces between software and hardware – are used, such as ROS (Robot Operating System). On the other hand, in hydrography, most software are under Windows. Managing communication between these two worlds is a real ordeal for our engineers!

Day 3

Today the weather was kind to us, windy but sunny. Some fearless students even went for a swim.

Focus 1 : Brave, the autonomous sailboat

The autonomous sailing boat named Brave was first designed by the IFREMER and built by ENSTA Bretagne to to compete in the World Robotics Sailing Championship (WRSC). This world champion autonomous sailboat is able to navigate thanks to a GPS and can change sail orientation by measuring and adapting to wind direction. Today third year students sailed it from one buoy to another using only camera detection. Tomorrow, they will test software to analyze images of buoys and will look for a way to detect them automatically while sailing.

Focus 2 : Can whales detect fishnets?

Some whales eat fish trapped in fishnets. To meet their quotas, fishermen end up casting more fishnets, which disturbs sea life.

Third years students are trying to understand how whales detect the presence of fishnets. To do so they immerse 4 hydrophones into the water to simulate the ears of whales.

With boats, they made noises at the surface by moving in the exact same way as fishing boats when they deploy fishnets. They managed to record the sounds of manoeuvres with different boats. Their next move is now to try and analyze and sort the recorded sounds using specific algorithms. This should help them determine whether whales can, just by hearing noises, detect the presence of fishing boats and fishnets.

This year is the first time second year robotics students have travelled to Guerlédan lake with the third year students. For us the aim of the stay is to work on GPS navigation with small boats called DDboats. These new boats were made by our robotics teachers Mr Benoit ZERR and his colleagues.

The stay is composed of a mixture of classes and practical work all aimed at reaching a point where the boats are able to perform a simple choreography on the lake without crashing into each other, a seemingly impossible feat in such short time.

Now the second day has come to an end and not a single DDboat has made it to the water. At that point most of the work has been done on computer simulations, which provide a simplified model of the boat and simulate the precision errors of the GPS. The hardest part is to estimate the position of the boats as precisely as possible.

Here is one example where the aim is to keep a given heading. The blue dots (which are mostly hidden by the other dots) correspond to the real trajectory of the boat. The red dots show the simulated GPS values, which have been randomly modified to fake the errors of real GPS measures. Finally, the green dots correspond to the estimated position. A system filters GPS values to get as close as possible to the real values.

The above picture shows the same kind of simulation but with a U-turn. Indeed, during the first attempts the simulated regulator was so unstable that a U-turn was impossible.

Finally, today was also the first time we had analyzed real world GPS data. Our teacher collected GPS data on the site, by walking the boat around. We than read this data and filtered it. The next picture is a plot of the real data (in blue) and the filtered data (in green). Finally the arrows correspond to the estimated speeds at each point.

And that ends the news for today, see you tomorrow for the latest updates !

Day 2

We’re back today with updates about our time at Guerlédan Lake. Here both the second- and third-year students are working on their respective projects while taking advantage of the lake for experiments.

Today we’ll focus on two groups of third year students and their respective projects.

The first group we shadowed are working on a sci-fi like project where the goal is to use a sensor called a LIDAR in cooperation with an inertial unit to locate yourself more accurately when a GPS is unable to access your position – because of obstacles or because you are underground.

Those students spent their day calibrating the LIDAR sensor (in the foreground of the picture) to make sure that the data it measures is correct. As it is humanely impossible to perfectly align the captor with the ground, this patch test ensures that the raw data won’t be affected by the offset. The LIDAR sensor is now able to compensate the flawed data.

In the meantime, the students working on the analysis of water columns in order to count fish populations were busy comparing and improving IFREMER’s and ENSTA Bretagne’s methods and equipment. The day was well spent diving into the arcane depths of the software as a first step towards their goal.

IFREMER’s software being incompatible with ENSTA Bretagne’s measuring protocol, they first needed to write a program that will enable them to draw charts of both data sets in order to compare them later this week.

Tune in for more news about exciting student projects tomorrow!

Day 1

After a two-hour bus trip, ENSTA Bretagne 3^rd year hydrography and oceanography students as well as 3^rd and 2^nd year robotics students discovered the inviting weather conditions at Guerlédan.

But why do ENSTA Bretagne students head to Guerlédan lake each year?

Two weeks a year, this place is turned into a real-size test laboratory. The lake bed was documented the last time the lake was emptied, so hydrography and oceanography students can explore it using the techniques they have learnt, and check their findings against the available data. The Guerlédan lake also provides students with rich opportunities to explore maritime application of robotics, conduct experiments and hone their problem-solving skills.

What did ENSTA Bretagne students do during this first day?

Once on site 3^rd year students quickly started working on their respective projects, while the 2^nd year robotics students attended a lesson on inertial navigation systems.

After lunch, students got back to work. 2^nd year students kick-started their English podcast production workshops by making contact with the groups they will be reporting on all week.

A focus on “Isobath tracking with an Autonomous Underwater Vehicle” project.

Four robotics students are currently working on a torpedo-shaped underwater drone.

This drone is meant to follow an isobath and must be autonomous as it is immerged.

An isobath is an imaginary line or a line on a map or chart that connects all points having the same depth below a water surface – an ocean, a sea, or a lake.

A thruster located at the back of the Autonomous Underwater Vehicle allows it to move forward. Four other water pumps steer the drone. The motor of the drone has been fixed today, as it was defective.

Focus 2: The BOATBOT

The boatbot project consists in automating a motorized zodiac so it can draw a map of the seabed. The zodiac can already steer itself but has yet to be able to regulate its speed, which is what Erwann, David and Philipert hope to achieve by the end of this week. They are also exploring ways to adjust the length of the underwater probe in order to avoid collisions.

Boatbot has been designed to look for La Cordelière, a Breton battleship built in the fifteenth century, equipped with two hundred guns and manned by a thousand crewmen. It was part of Duchy of Brittany’ Navy and was destroyed during the battle of Saint-Mathieu: while the ship was in Brest hosting the celebrations of Saint-Laurent’s day, the English fleet was preparing a landing in Brittany, and the Cordelière was dispatched to intercept them. It was met by a larger English fleet, and fought valiantly against the British ship “Regent”, and both ships were destroyed after the Regent tried to board the Cordelière. It is estimated that two thousand men died in the battle, and among them were three hundred civilians still on board the Cordelière. All previous attempts to locate its wreck have been in vain and its location remain unknown in Brest bay.