List of winning projects
Theses:
Project title: Recognition and interpretation of the environment for safe autonomous navigation and differentiated action in agricultural settings
The use of robotic tools to support the agroecological transition requires a high level of autonomy, guaranteeing on the one hand the ability to act in a targeted and discriminated manner, while ensuring operational integrity. These characteristics are fundamental for these technologies to be efficient and acceptable to farmers in supporting their new practices. Such a level of autonomy and precision requires robots to be able to interpret the environment reliably in order to adapt their behavior to the context and the agronomic tasks to be performed, and to assess the risk of crossing the areas to be treated. While this concept is well understood in indoor robotics, it is more delicate in agricultural environments due to the changing nature of the environment and the layout of the plots. It therefore appears necessary to develop local semantic mapping tools that can distinguish the type of environment, its passability, and the type of operation permitted in different areas. The aim of this thesis will therefore be to develop environment recognition tools using multi-sensor fusion. Based on the definition of an ontology integrating agroecological and robotic constraints, the aim will be to propose segmentation and learning approaches to define labeled 3D spaces. These will be used to modulate the robot's behavior for safety purposes (speed reduction, prohibition), or for specific treatments (differentiated weeding, picking, etc.), by adapting sensory-motor behaviors. The results of this thesis will enhance the autonomy and efficiency of robots in performing specific operations on crops, while ensuring the safe movement of autonomous vehicles in a changing environment.
Project title: AphiLeg, towards an explainable multimodal model for predicting the resistance of legumes to insect pests
Legumes such as peas and field beans are key species for agroecological transition. However, while legumes benefit from significant genetic and genomic resources, their detailed characterization for pest resistance remains a major challenge. The AphiLeg project aims to develop innovative, explainable, multi-modal methodologies to predict the resistance of legumes to aphids, the main insect pests of these crops. By combining digital vision, genomics, and metabolomics, this project proposes a multi-modal approach to improve resistance prediction and provide interpretable explanations of the underlying mechanisms. AphiLeg will provide a better understanding of plant-insect interactions by providing new biological hypotheses derived from modeling the interactions between genotype, phenotype, and metabolism. This methodology will also be a relevant decision-making tool that can be used in pre-breeding and varietal selection. In the long term, this project could have a significant impact on sustainable crop management and reducing pesticide use.
Project title: Multi-agent modeling of the combined effects of farming practices and earthworms on soil structure and proposal of soil health indicators
- PhD student: Amandine Ouédraogo
- Affiliated unit:
- Co-supervision: Cécile Chéron-Bessou (IRD) and Nicolas Marilleau (IRD/UPMC)
- Doctoral school: Doctoral School GAIA
- Project duration: 2025-2028
This thesis aims to contribute to soil preservation by developing indicators of the impact of agricultural practices on soil health within the framework of life cycle assessment (LCA). Currently, LCA does not take soil health into account in a conceptually robust manner. To remedy this, the research draws on the framework developed by Kibblewhite et al. (2008), which defines soil health according to four biological functions. The study focuses in particular on the role of earthworms in maintaining soil structure, as they have a major impact depending on their lifestyle (anecians, endogeans, decompactors, or compactors). However, experimental study of these interactions is costly and sometimes unfeasible, which justifies the use of theoretical and mechanistic models. Multi-agent models, such as SWORM and CAMMISOL, can be used to simulate interactions between earthworms, soil structure, and farming practices. These models still need to be improved and integrated in order to establish a robust tool capable of translating the links between agricultural practices and soil condition.
Project title: Development and implementation of learning approaches for the study of RNA modifications related to climate change adaptation in Arabidopsis thaliana
- PhD student: Emma Rodriguez
- Affiliated unit:
- Co-supervision: Christine Gaspin (CNRS) and Julio Saez-Vasquez (CNRS/UPVD)
- Doctoral school: Doctoral School SEVAB (Ecological, Veterinary, Agronomic and
- Bioengineering Sciences)
- Project duration: 2025-2028
Epitranscriptomics, defined as the set of RNA modifications (>170 listed modifications), introduces an additional layer of complexity into the study of gene expression regulation and translation. Numerous studies have highlighted the importance of these modifications in several essential biological processes. Nanopore's Direct RNA technology is promising because it theoretically allows access to all modifications at the level of the entire epitranscriptome. In particular, it has confirmed the major impact of chemical modifications of RNA in the field of health. Several methodological barriers remain to be overcome in order to make the most of this technology. We propose to overcome some of these obstacles by developing deep learning methods. As part of this thesis, we propose to develop/improve analysis methods using public and generated data on the model plant Arabidopsis thaliana to identify epitranscriptomic signatures associated with stress tolerance (thermal and water).
Post-doctoral fellowships:
Project title: CogniBot: A prototype for autonomous and personalized cognitive enrichment for pigs
- Postdoctoral researcher: Louise Kremer
- Affiliated unit:
- Co-supervision:
- Project duration: 2025-2027
Project title: Machine learning and agroecological transition: digital social networks and professional socialization of agricultural high school students
- Postdoctoral researcher: Laurianne Trably
- Affiliated unit:
- Co-supervision: Sylvain Brunier (INRAE/AgroParisTech) and Corinne Robert (CERES-ENS)
- Project duration: 2025-2027
The climate emergency and environmental impacts make agriculture and farming models a central focus of research. The younger generations are a key link in the transition to more virtuous practices, but their adherence to the agroecological practices taught in schools remains limited. This project focuses on how content related to agroecology on the internet is received and perceived as legitimate (or not) by young farmers training in vocational high schools. It seeks to understand the effects that content circulating on social media may have on the professional practices of young people in training, and how they contribute to defining the scope of possible agroecological practices, beyond the knowledge transmitted by school or family alone. To answer this question, a qualitative survey (interviews, observation, gray and specialized literature) and a quantitative survey (analysis of online comments, questionnaires) will be conducted among young men and women in three agricultural high schools in metropolitan France, thanks to close interdisciplinary collaboration between sociology and agronomy.
