Fully funded PhD in security of IA on drones, starting from Nov/2025
Full description available here: https://sdrive.cnrs.fr/s/XHyjcGtEPHemZSF
Embedded distributed AI for collaboration within a Swarm of Drones
Scientific context
This PhD is in the frame of the ANR Junior Professor Chair on Autonomous and Secure Swarm of Maritime
Drones. The Chair aims at providing trust and security on maritime swarm of drones, a strategic and major
issue in the current socio-political context.
The usage of drones, aerial or maritime (surface or underwater) is becoming more and more widespread in var-
ious domains, recreational and critical. Drones, more and more intelligent and autonomous for their missions
and navigation, embed more and more a certain form of Artificial Intelligence (AI). However, drones are em-
bedded systems with limited resources and energy. One solution for deploying the very large neural networks re-
quired for certain complex missions on these constrained systems is to distribute the model among several collaborative drones. In this solution, the model and data are distributed across several partitions, and each parti-
tion is implemented in a different device. Various efforts have been made to design efficient partitioning strategies
that optimise different criteria such as network performance or energy consumption. However, partitioning
also depends on the nature of the task and on the resources available in a swarm of drones, which can be
heterogeneous and include drones of different sizes andcomputing resources.
In this thesis, we focus on the on-board deployment of distributed intelligence for the collaborative detection
of deviant behaviour within a swarm of drones. For this complex task, different software and hardware signals
and information are collected at different levels within the swarm the swarm, and can be processed locally and
collaboratively. Finally, depending on the resources available, the model can be reconfigured among the devices
to optimise resource use and activity within the swarm.
Objectives and Work Program
This PhD work will focus on the efficient embedded deployment of distributed, resilience, artificial intelligence models within swarm of drones. The targeted application will be the collaborative detection of deviant drone behaviour, an application which will require the dynamic consideration of resource availability within the pack for the efficient distribution of the model.
- The main objectives of this PhD are as follow:
The study and design of novel approaches based on machine learning techniques for the collaborative
detection of deviant behaviour within a swarm of drones. Detection will be considered at several levels,
both centrally within a mothership-type drone with larger resources, and in a hybrid distributed manner
between heterogeneous drones. For this task, the student will benefit from the experience of the Lab-
STICC’s SHAKER team in implementing on-board artificial intelligence algorithms. - A study, depending on the level under consideration, of efficient techniques of distributing the model
between drones with different computing and memory resources. - The efficient design of techniques for dynamically distributing the model efficiently between the various
devices and resources available. - Depending on the progress of the work, we will study the robustness of the implemented distributed
model when facing denial-of-service attacks such as adversarial attacks that could compromise the system’s
detection result. - The prototyping and evaluation of the design solutions will be carried out via the implementation on
heterogeneous systems-on-chip mastered by the team, e.g. SoCs (ARM+GPU), and FPGA boards
(ARM+DPU on FPGA), and will benefit of the drones platforms and expertise of Lab-STICC labo-
ratory, Lorient. Validation will be in terms of performance, including model distribution time, and energy
consumption. In addition, in collaboration with a second thesis in progress on the subject, we will evaluate
the effectiveness of deviant behaviour detection (accuracy, f1-score, etc.).
Candidate Profile
Graduate: candidate must hold a Master’s degree or an Engineer degree in Computer/Electrical Engineering,
Computer Science, Embedded Systems, or similar. The required skills are:
- Interest in, familiarity with and/or good knowledge of artificial intelligence, model generation tools and
deployment on processors, FPGAs and/or GPUs. - Good knowledge in C/C++/Python
- Good knowledge in Embedded Systems and Computer Architecture
- Cybersecurity knowledge, skills are not mandatory, but interest in this topic is required
Date and place
- Applications are open until end of April.
- Although flexible, this PhD is expected to start from October 2025 for a 3-year duration.
- The PhD student will work within SHAKER and ARCAD teams of the Lab-STICC laboratory in Lorient, France and will benefit of SHAKER team drone expertise and platforms, as well as of the ARCAD team
expertise in cybersecurity.
How to apply
Please send your CV and available Master/Engineering transcripts to:
- Maria Méndez Real - maria.mendez-real@univ-ubs.fr
- Vianney Lapôtre - vianney.lapotre@univ-ubs.fr
- Dominique Heller - dominique.heller@univ-ubs.fr