Environmental analysis and modelling

New technologies and associated methods can be used to analyse and reduce environmental impacts. Agricultural equipment, robotics and information technology play an important role in the design of tools for the agro-ecological transition.

In this context, perception, analysis and modelling of the environment are essential. The aim of this research group is to design and experiment with new approaches to these issues. One of the circle's roles is to enable interaction between different engineering science disciplines, all of which are concerned with the perception, analysis and modelling of the environment. The research carried out is on a relatively local scale, ranging from a territory to a plant.

The environment can be perceived by interactions with or without contact: mechanical, electrical, laser, camera, radar, etc. The information can be analysed using digital methods and technologies. The circle is thus interested in various interconnected subjects:

• Environment recognition methods using contact interaction

This topic concerns the measurement of deformations, interaction forces and other physical properties. For example, an agricultural robot can use mechanical sensors to determine the traversability of an object. This will make it possible to measure the deformability of obstacles and thus differentiate between a non-deformable object (a wall) and a deformable object (tall grass). Another example of a mechanical sensor is a tire fitted with sensors to characterise soil properties (compaction, moisture, etc.).

• Non-contact environmental perception methods

Typically, laser, camera (photogrammetry, TOF), microwave radar, multispectral and hyperspectral technologies are at the heart of our research. The systems we develop are deployed on a variety of platforms, from land vehicles to aerial drones. One example of an application concerns images acquired by drone. Photogrammetric processing can be used to build a 3D model of the environment overflown. This model provides ground robots with a priori knowledge of the environment in which they will be operating.

• Digital technology methods for processing and modelling environmental data

These include artificial intelligence techniques, signal processing, statistics and environmental data modelling. Artificial intelligence makes it easier to recognise obstacles, for example. Another facet concerns geographical data modelling methods. Various methods are being developed for specifying data, including extensions to formalisms such as UML (Unified Modelling Language). These different methods will also be used to design digital twins for modelling agricultural equipment.

Click here to download the full presentation

Some examples of publications:

Sensitive Device for Probing and Recognition of Obstacles in a Natural Environment, Lama Al Bassit, Hawraa Becher, Bastien Laurent, Hubert Villeneuve, Guillaume Jeanneau. Agritech Day, Axema, Oct 2023, Rennes (FR), France. pp.66-75

Risk-Aware Navigation for Mobile Robots in Unknown 3D Environments, Elie Randriamiarintsoa, Johann Laconte, Benoit Thuilot, Romuald Aufrère, 2023 IEEE 26th International Conference on Intelligent Transportation Systems (ITSC), Sep 2023, Bilbao, France. pp.1949-1954, ⟨10.1109/itsc57777.2023.10420846⟩

Measuring the behaviour of lambs in an isolated environment with artificial intelligence methods,Bernard Benet, Romain Lardy, International congress on animal science EAAP / WAAP Conference 2023, EAAP / WAAP, Aug 2023, Lyon, France. pp.970

Robot localization and navigation with a ground-based microwave radar, R. Rouveure, C. Debain, R. Peuchot, J. Laneurit. International Radar Conférence (RADAR 2019), Sep 2019, Toulon, France