(M/F) PhD position : Foundation Models for Interaction Modeling in Autonomous Mobile Robotics

Job Description

Thesis Subject

Research Context
Recent advances in mobile robotics have been driven by remarkable progress in perception, deep learning, and control. However, current robotic systems still face major challenges when operating in fluid environments such as air or water. Unlike ground robots, aerial and underwater robots exhibit dynamics that are strongly coupled with their surrounding environment through complex aerodynamic or hydrodynamic interactions. These interactions are highly nonlinear and depend on the robot geometry, local flow conditions, turbulence, and external disturbances.
Current control methods generally rely on simplified interaction models based on constant aerodynamic coefficients, quasi-static approximations, potential flow models, or experimentally identified empirical models. While these approaches are adequate for basic stabilization tasks, they become insufficient in more challenging scenarios such as flight near obstacles, drone swarms, wind interaction, navigation in waves or ocean currents, and underwater manipulation.
At the same time, Foundation Models have transformed several fields of artificial intelligence by learning general-purpose representations from massive datasets. Their ability to capture transferable knowledge offers a unique opportunity to rethink the modeling of physical interactions. The central idea of this PhD is that a model trained on large-scale numerical simulations, instrumented experiments, and onboard robotic data could learn generic representations of local interactions between mobile robots and their environment, enabling transfer across different robotic platforms and morphologies.
More generally, the dynamics of a robot can be represented by a generic differential equation describing the evolution of its state (x) as a function of control inputs (u), external disturbances (p), and interaction terms (d). Depending on the application, these interaction terms d may represent aerodynamic or hydrodynamic forces, interactions between multiple robots, contacts with manipulated objects, or other environmental effects. In conventional approaches, these interactions are typically described using highly simplified analytical models, for example, assuming thrust to be proportional to rotor speed in aerial vehicles.
In multi-robot systems, the interaction terms encompass a much broader range of phenomena, including aerodynamic coupling between drones, perception and collision avoidance, communication effects arising from information sharing, and mechanical interactions during cooperative payload transportation or through physical connections between robots.
The central hypothesis of this PhD is that these interaction terms can instead be represented by a generic Foundation Model capable of capturing local physical interactions across multiple robotic platforms. This model would leverage information about the robot morphology, geometry, internal structure, state, control commands, external disturbances, and environmental conditions to predict interaction effects. Unlike robot-specific neural network models, the proposed approach aims to learn a universal representation of local interactions (fluid-structure, robot-robot, robot-object, etc.) that can be efficiently adapted to new robotic platforms using only a small amount of additional data through few-shot adaptation.
Research Objectives
The primary objective of this PhD is to develop a new generation of hybrid models combining large-scale machine learning with physical knowledge to represent interactions between mobile robots and their environment.
The research will address several fundamental scientific questions:
• Which geometric representations are best suited for learning transferable physical interactions?
• What types of data are required to train such a generic interaction model?
• How can strong generalization across different robotic platforms and environments be achieved?
• How can these learned interaction models be integrated into stable and reliable feedback control systems?
The proposed methodology will be investigated in several application domains, including: • Fluid-structure interactions for aerial and underwater robots.
• Cooperative transportation of payloads by multiple drones.
• Coordinated swarm and formation flight with interaction-aware control.
This PhD will be carried out within a national research project dedicated to the development of next-generation digital technologies for agriculture and environmental sciences. The research will contribute to new capabilities for autonomous observation, monitoring, and data acquisition using intelligent robotic systems.
Potential application areas include precision agriculture, crop monitoring, biodiversity assessment, wildlife observation, environmental monitoring, and the study of pastoral and ecological systems.

Requirements
• M.Sc. degree in robotics, control engineering (or related fields)
• Excellent written and spoken English skills
• Good experience in C/C++, ROS, Matlab/Simulink
• Previous research experience, particularly in the context of a Master's thesis or research internship, would be highly valued.
• Scientific curiosity, autonomy, and ability to work independently
• Knowledge of, or experience with, learning methods for robotics would be a plus
• Experience with robotic systems and/or aerial robots is a plus

Your Work Environment

Gipsa-lab is a joint research unit of the CNRS (French National Centre for Scientific Research), Grenoble-INP (Grenoble Institute of Technology), and the University of Grenoble, affiliated with Inria (French National Institute for Research in Digital Science and Technology) and the Grenoble Observatory of Earth Sciences.

With 350 people, including approximately 150 doctoral students, Gipsa-lab is a multidisciplinary research unit conducting both fundamental and applied research on signals and complex systems.
Gipsa-lab develops projects in the strategic fields of energy, the environment, communication, intelligent systems, life and health technologies, and language engineering.
Through its research activities, Gipsa-lab maintains a constant link with the economic environment thanks to strong partnerships with businesses.
Gipsa-lab staff are involved in teaching and training at various universities and engineering schools in the Grenoble metropolitan area (Université Grenoble Alpes).
Gipsa-lab is internationally recognized for its research in Automation and System Safety, Data Science (Information and Signal Processing), Speech, and Cognition. The unit conducts its research through 11 teams organized into 4 research departments:
- Automation and System Safety
- Data Science
- Geometry, Learning, Information and Algorithms
- Speech and Cognition
The Gipsa-lab comprises 150 permanent staff and approximately 250 non-permanent staff (doctoral students, post-doctoral researchers, visiting researchers, master's students, etc.)

Constraints and risks

NONE

Compensation and benefits

Compensation

2300 € gross monthly

Annual leave and RTT

44 jours

Remote Working practice and compensation

Pratique et indemnisation du TT

Transport

Prise en charge à 75% du coût et forfait mobilité durable jusqu’à 300€

About the offer

About the CNRS

The CNRS is a major player in fundamental research on a global scale. The CNRS is the only French organization active in all scientific fields. Its unique position as a multi-specialist allows it to bring together different disciplines to address the most important challenges of the contemporary world, in connection with the actors of change.

CNRS

The research professions