Open Master/Engineering internship Position on “Federated learning for mobile user localization based on VLC in dynamic indoor environments“
13 Février 2023
Catégorie : Stagiaire
Open Master/Engineering internship Position on
“Federated learning for mobile user localization based on VLC in dynamic indoor environments “
A project supported by the European H2020 Research project: 6G BRAINS
Starting date:April 2023 (Negotiable if necessary) for 5 months
Most of IoT applications such as emergency services, e-marketing and social networking critically exploit location information to operate properly. Traditional indoor localization methods are affected by multipath effects caused by the high dynamic of indoor environments which can lead to high localization errors. To address this problem, machine learning techniques have been considered for data-based localization. Such promising learning models rely on data collected offline to capture variations in indoor environments. Localization methods based on conventional machine learning models entail the collection of data from IoT devices into a central server/unit which results in a lot of data exchange with the server, privacy concerns and a high reliance on the server. This requires a high bandwidth and assumes the server to be trustworthy. This may lead to issues in practical considerations where data are distributed across IoT devices in a privacy-preserving objective. Consequently, federated learning (FL) was developed to conserve bandwidth while protecting the privacy of users’ data. Indeed, FL is becoming more attractive for localization problems with its great advantage of privacy, bandwidth optimization which is promising in indoor localization field especially when dealing with high dynamic environments which do require a frequent model update due to the continuous change of propagation conditions.
In this context, we propose to design a learning scheme for 2D coordinates prediction in one floor environment based on DNN. Then, regarding the resource constraints of IoT devices, we propose a federated learning framework to train the proposed model, yielding a communication-efficient collaborative and privacy-preserving indoor localization. At first, the proposed scheme is validated based on simulated data generated using existing VLC models on Matlab. Once validated, it will be tested on real data collected on ISEP laboratory in order to strengthen simulation results.
Who should apply: This internship has a minimum 5 months duration beginning in April 2023. Internships will be awarded on a rolling basis and candidates are encouraged to apply early.
-A master's student in computer science, or related field (exceptional undergraduates will also be considered).
-Sufficient knowledge in the field of AI and machine learning.
-Experience building systems based on machine learning and/or deep learning methods.
-Research experience demonstrated via an internship, work experience, or coding competitions.
-Knowledge in Python and/or Matlab.
-Strong communication skills and teamwork experience.
-Good level in oral and writing English (French optional).
-You will work in multidisciplinary environment.
-You will be part of an international team.
-You will Influence progress of relevant research communities by producing publications.
-You will have access to state-of-the-art testbed implemented at ISEP.
-You will work on a European project (6GBrains: https://6g-brains.eu/).
-Upon completion of the internship and based on the results, you will be offered a 3 years grant to pursue a PhD at ISEP.
 YIN, Feng, LIN, Zhidi, KONG, Qinglei, et al. FedLoc: Federated learning framework for data-driven cooperative localization and location data processing. IEEE Open Journal of Signal Processing, 2020, vol. 1, p. 187-215.
 ETIABI, Yaya, CHAFII, Marwa, et AMHOUD, El Mehdi. Federated Distillation based Indoor Localization for IoT Networks. arXiv preprint arXiv:2205.11440, 2022.
Application procedure: Applications should include the following elements: 1) a detailed CV, 2) Two academic references, 3) a one-page Motivation Letter.
All applications must be submitted directly by email at email@example.com and firstname.lastname@example.org . For applications, the subject of your email should be “Application for FL-based localization internship”.
Closing date: 11:59 pm 22th February 2023 (GMT+1 Time Zone)
Interviews will be conducted by videoconference between 23-24February 2023.