Practical information

• Object: The RIS team at LAAS-CNRS in Toulouse (France) is searching for a Master's student interested in performing his/her internship.
• Location: LAAS-CNRS, Toulouse, France.
• Starting date (flexible): Spring/Summer 2026.
• Duration: 4-6 months.
• Salary: about 650 EUR net per month.

Links to the internships

https://partage.laas.fr/index.php/s/LKKGeTW8GBr33Wf

Scientific project

Summary

The purpose of this project is to integrate machine learning techniques with optimal control strategies to design safe and intuitive frameworks for human–robot (e.g., human-drone) collaboration. The focus will be on the collaboration between human operators and robots, exploring how adaptive algorithms can capture human intent and translate it into coordinated robotic actions. By combining learning and control, this work seeks to enable more natural and robust interaction scenarios.

Project description

A promising strategy towards effective collaboration lies in imitating the way humans cooperate with each other. An important element in human collaboration is the capacity to predict the collaborator’s actions and act accordingly. If robots were able to anticipate human intentions (such as future movements or expected contact forces [1]), they could operate more safely and efficiently around them. Artificial intelligence, and in particular deep learning, provides powerful tools to achieve this goal. By integrating predictive models (such as Transformer-based neural networks [2]) into optimal control paradigms, it becomes possible to design interaction strategies that are both safe and human-like.

Several complementary research directions can be pursued depending on the student’s interests. For each direction, we indicated the academic backgrounds that appear most relevant; however, these are only indicative, and we welcome students from diverse educational paths:

• [Computer science/Robotics] Safe Learning for Control. Deep learning has proven highly effective in solving complex control problems, yet guaranteeing stability, safety, and constraint satisfaction remains an open challenge. Recent approaches, including Physics-Informed Neural Networks, Bayesian Neural Networks, and stability-constrained neural architectures, aim to make learning-based models more reliable and explainable. This research direction focuses on designing and evaluating learning-based predictors of human actions that can be safely integrated into control loops.
• [Robotics] Shared autonomy in human–robot interaction. Investigation of shared autonomy [3], a novel human-robot interaction paradigm where the level of robot autonomy is dynamically adapted based on the robot’s understanding of human intentions and the surrounding environment. In this way, the robot becomes capable of adapting to human behavior. This approach holds particular promise for improving safety, efficiency, and transparency in human–robot collaboration.
• [Biomedical Engineering/Robotics] Development of human assistive algorithms. As Industry 5.0 places strong emphasis on human-centered automation, collaborative robots are increasingly expected to improve not only productivity but also human comfort, safety, and physical well-being. In this context, we aim to extend our current framework (Our framework aims to improve ergonomics in repetitive collaborative tasks by combining short-term human hand trajectory prediction with an MPC that generates supportive robot motions while preserving user intent.) toward assistive algorithms that promote smoother, more intuitive, and ergonomically comfortable human–robot interactions over large workspaces. The integration of force sensing and physiological signals (e.g., EMG) will enable adaptive control strategies that modulate robot behavior based on the human’s physical state, effort, and intention.
• [Computer science] Transfer learning across robotic platforms. Exploration of transfer learning strategies, where models trained on data collected from a ground manipulator are adapted and deployed on aerial robots.

Project environment

The student will join the “Robotics and InteractionS” (RIS) team, being one of the three robotic equipes within LAAS-CNRS in Toulouse. RIS is an internationally recognized research team focused on developing autonomous mobile machines that integrate perception, reasoning, learning, action, and reaction capabilities. The main research areas of the team are Architectures for Autonomous Robots, Learning, Temporal Planning and Execution, and Algorithmic Motion Planning and Control. RIS is composed of 11 permanent researchers, 3 PostDocs, and several Ph.D. students.

More generally, research at LAAS-CNRS spans robotics, optimization, control, telecommunications, and nano-systems. The robotics department at LAAS-CNRS counts more than 100 people, and it is one of the largest and oldest robotic research departments in France.

The LAAS-CNRS robotic department has made world-class contributions in artificial intelligence, planning, perception, and autonomous robotic systems, such as humanoids and aerial robots.

Useful links:

• RIS webpage: https://www.laas.fr/public/en/ris
• LAAS-CNRS homepage: https://www.laas.fr/public/en

Candidate profile

• [Essential] Bachelor's degree in robotics, engineering, applied mathematics, biomedical, computer science (or related fields)
• [Essential] Very good skills in C/C++ and Python, and, in general, the ability to independently manage large software projects
• [Essential] Scientific curiosity and a “thirst for knowledge” approach to learning
• [Desired] Experience with Machine Learning and Neural Networks, ideally knowledge of common Machine Learning Python libraries such as PyTorch or TensorFlow
• [Desired] Solid mathematical background (ideally in robotics).

Supervisor’s profile

The main advisor is Dr. Marco Cognetti (https://scholar.google.com/citations?user=NAq-dYcAAAAJ). The student will be co-supervised by Alessia Fusco, a PhD student at LAAS-CNRS.

How to apply

Interested candidates are requested to apply by sending their CV and a motivation letter to and with the subject [Internship LAAS - shared autonomy]. The selection procedures will consist of an interview that will be organized in English. The candidate will be asked to present his/her previous experience and to discuss the project. The candidate will also be asked to solve a small exercise to assess his/her mathematical and programming skills. The position will remain open until satisfactory candidates are found.

Internship duration/funding/Ph.D. opportunity

The internship will (ideally) last 6 months, and each student will receive about 650 EUR net per month as reimbursement of expenses. The internship can lead to a Ph.D. in the RIS team if both the student and the supervisor are interested. Starting internship date (flexible): spring/summer 2026.

References

[1] M. Selvaggio, M. Cognetti, S. Nikolaidis, S. Ivaldi, and B. Siciliano, “Autonomy in physical human-robot interaction: A brief survey,” IEEE Robotics and Automation Letters, vol. 6, no. 4, pp. 7989–7996, 2021.

[2] A. Fusco, V. Modugno, D. Kanoulas, A. Rizzo, and M. Cognetti, "Transformer-Based Prediction of Human Motions and Contact Forces for Physical Human-Robot Interaction," 2024 IEEE International Conference on Robotics and Automation (ICRA), Yokohama, Japan, 2024, pp. 3161-3167, doi: 10.1109/ICRA57147.2024.10611211.

[3] A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, Ł. Kaiser, and I. Polosukhin, “Attention is all you need,” Advances in neural information processing systems, vol. 30, 2017.

[4] Jessica Cauchard, Charles Duta, Gianluca Corsini, Marco Cognetti, Daniel Sidobre, et al. Considerations for Handover and Co-woruking with Drones. HRI ’24: ACM/IEEE International Conference on Human-Robot Interaction, Boulder, CO, USA.