Description

This PhD thesis aims to develop novel visual servoing methods for robotic tasks such as assembly, inspection, or welding. The challenge and originality of the topic lie in addressing industrial production constraints identified by the members of the IRT Jules Verne:
• Lack of precise a priori calibration (task confinement, prohibitive cost of metrology),
• Absence of training data (small batches or unique parts),
• Potential complexity of trajectories.
Your mission will be to simultaneously address these challenges by revisiting, generalizing, and extending a method previously tested in a past project. The core principle of the method involves generating vector fields in the camera image space, integrating task constraints and obstacles. The actuation system is then velocity-controlled using the vector field, enabling behavior control without explicit temporal dependence.
The thesis will tackle systems of increasing complexity, starting with minimal theoretical examples in reduced dimensions, then progressing to more complete simulated and real-world systems.

Key Activities
• Targeted literature review on visual servoing, with a focus on the stability analysis of control laws,
• Mathematical study of control laws,
• C++ development using ROS 2, OpenCV, and OpenMesh frameworks/libraries,
• Experimental validation on a setup comprising a cobotic arm and one or more cameras (at ICA Albi),
• Deployment of the solution on at least one industrial demonstrator at IRT Jules Verne,
• Scientific communication (conferences, journals) and manuscript writing (in LaTeX).

Skills required

You hold a Master’s degree in Robotics, Control Engineering or Applied Mathematics, with a strong aptitude for mathematics and their practical implementation through programming and experimentation. Required Theoretical Knowledge: • Mathematics: Linear algebra, differential calculus, numerical methods, • Computer Science: Algorithms, programming (preferably C++), complexity analysis, • Control Engineering: Fundamentals (linear control, frequency analysis, PID), • Robotics: Geometric, kinematic, and dynamic modeling (basics). Expected Technical Skills • Mathematical analysis and modeling of problems, • Development of complex software projects with third-party library integration, • Rigorous scientific experimentation methodology, • Communication skills: Ability to synthesize work, report progress, and clearly articulate challenges, • Familiarity with ROS middleware and OpenCV libraries (desirable but not mandatory).

Bibliography

• X. Sun, X. Zhu, P. Wang and H. Chen, A Review of Robot Control with Visual Servoing, 2018 IEEE 8th Annual International Conference on CYBER Technology in Automation, Control, and Intelligent Systems, 2018.
• G. Chabert, F. Chaumette, A. Suarez Roos. Clevis and Tenon Assembly Using Visual Guiding Fields, IROS 2025 - IEEE/RSJ International Conference on Intelligent Robots and Systems, 2025.
• J. Pomares and F. Torres, Time Independent Tracking Using 2D Movement Flow-Based Visual Servoing, IEEE Int. Conf. on Robotics and Automation (ICRA), 2005.
• R. Reyes, I. Becerra and R. Murrieta-Cid, Visual-RRT*: Generating Asymptotically Optimal Trajectories With Vision-Based Controllers, IEEE Transactions on Control Systems Technology, 2026

Keywords

Visual servoing, Process robotics, Robotic arm