This thesis will make it possible to discover the mechanisms of the telecommunications networks which will integrate the 6G of tomorrow with a focus on drones for the user-equipments. In addition, it will be an opportunity to design and implement so-called artificial intelligence algorithms that make it possible to define optimal approaches for learning, decision support, and traffic prediction of connected drone movements.
5G Networks; 6G Networks; Drones ; Machine Learning; Predictive Algorithms; Graph Neural Networks.
UAV (Unmanned Aerial Vehicles) are mobile stations that can fly to any area without human pilot intervention. Drones can be also remotely piloted. Initially, they were launched for military applications but later on, due to low cost, adaptive communication, and several other useful features made it an important breakthrough technology in civilian uses like cargo delivery, wireless communication, search and rescue, and precision agriculture . UAV-assisted heterogeneous networks represent a cost-effective solution to manage several use cases like sudden communication infrastructure failures, communication malfunctioning, or a sudden increase in number of users demanding connectivity and high data rate . For instance, in , the work proposes UAV-assisted backscatter communications, which combines the advantages of both UAVs and backscattering communications. A UAV acts as a data collector from multiple terrestrial backscattering tags via time division multiple access (TDMA). UAV communication is also considered as an emerging technology for 5G networks and beyond. On the one hand, UAVs can be quickly and efficiently deployed to support existing cellular networks and enhance ground users’ quality-of-service (QoS) . The use of drones is booming, especially in areas such as surveillance and delivery. As such, the mobile communications ecosystem (eg: , , ) studies the needs and impacts of these drones in terms of cellular connectivity. The impacts are still difficult to predict due to specificities due to drones (e.g. movement at altitude), but machine learning techniques allow increasingly advanced modeling and estimations  offer the opportunity to try to answer this challenge.
To date, the majority of drones use WiFi technology to connect to an internet network. Nevertheless, some drones offer the possibility of using the cellular networks of mobile telephone operators, which is extremely interesting from the perspective of moving drones over long distances. However, to date, the impact of such drones on operator networks has been little studied.
The objective of the thesis is to model and evaluate the impact of trajectories of connected drones on a radio access network, taking into account the roads and the cellular coverage of one or more operators. The purpose of the model will be to take into account planned drone routes to 1/ minimize the impact in terms of traffic on a given RAN (exclusively Orange RAN, or multi-MNO RAN, etc.), 2/ avoid interference between drones and sensitive areas, 3/ maintain good QoE for drone-type User-Equipment (UE) as well as for classic user-type UEs.
The main stages of this Ph.D. thesis are:
Study of the learning of neural networks on graphs (Graph Neural Networks, GNN) , which can be used to model and make predictions of drone traffic during their movement on the network of mobile telephone operators. It will be necessary to study the features essential to the quality of a connection between the drone-type user-equipment and the antenna of an operator: the characteristics relating to 3GPP cellular technology [10, 11], but also those essential regarding geographical clues, or even essential features related to geographic information systems (GIS). It will also be necessary to study the modeling of the mobile access network (Radio Access Network, RAN) of an operator, more particularly the geographical footprints of the radio cells of this RAN. A model could be based on a GNN type network. The whole system (data and model) will offer a compromise between an interesting complexity and a realistic computation time in the context of deep learning type implementation .
Several models will be tested in a single-operator context. The results will allow recommendations related to the trajectories of the drones in terms of geographical routing, but also in terms of day and time at which the connectivity will be the best for the drones. Depending on the progress of the research, the first results must be valued in the form of publications and/or patents. The models can be extended to take into account a multi-operator framework where collaboration between mobile operators would allow better connectivity over the entire trajectory of a drone.
Writing of the thesis manuscript and the valorization of the results with well-ranked publications and/or patents.
The candidate must hold an MSc diploma (national Master's degree) or another degree conferring the French Master's degree (Master 2). It is very preferable that the candidate has a good background in the field of machine learning, and ideally in the sub-fields of reinforcement learning, and/or techniques/software used in the field of trading.
The following qualities/skills are also considered:
- Knowledge of mobile networks,
- Deep Learning (PyTorch, Tensorflow or JAX)
- Graph theory
- Good communication in English
- Team work Curiosity, open-mindedness
 Z. Yong, R. Zhang, and T. J. Lim. "Wireless communications with unmanned aerial vehicles: Opportunities and challenges." IEEE Communications magazine 54.5 (2016): 36-42.
 R. Shahzadi, M. Ali, H. Z. Khan, M. Naeem,
UAV assisted 5G and beyond wireless networks: A survey, Journal of Network and Computer Applications, Volume 189, 2021, 103114, ISSN 1084-8045.
 S. Yang, Y. Deng, X. Tang, Y. Ding and J. Zhou, "Energy Efficiency Optimization for UAV-Assisted Backscatter Communications," in IEEE Communications Letters, vol. 23, no. 11, pp. 2041-2045, Nov. 2019.
 R. Duan, J. Wang, C. Jiang, H. Yao, Y. Ren and Y. Qian, "Resource Allocation for Multi-UAV Aided IoT NOMA Uplink Transmission Systems," in IEEE Internet of Things Journal, vol. 6, no. 4, pp. 7025-7037, Aug. 2019.
 Exploring future opportunities for mobile networks and the drone industry, use-case 5, 2021, https://www.gsma.com/iot/wp-content/uploads/2021/03/Drone-Benefits-Anal…
 GSMA, "Interface for Data Exchange between MNOs and the UTM Ecosystem", 2021, https://www.gsma.com/iot/wp-content/uploads/2021/02/ACJA-NetworkCoverag…
 3GPP "New SID on Additional capabilities of mobile networks for drone operations and management", China Mobile, November 2021, https://www.3gpp.org/ftp/tsg_sa/WG1_Serv/TSGS1_96e_EM_Nov2021/Docs/S1-2…
 "ETA Prediction with Graph Neural Networks in Google Maps", 2021, https://arxiv.org/pdf/2108.11482.pdf
 “Geometric Deep Learning: Grids, Groups, Graphs, Geodesics, and Gauges”, Michael M. Bronstein, Joan Bruna, Taco Cohen, Petar Veličković”, 2021, https://arxiv.org/abs/2104.13478
 3GPP TS 36.214 “LTE; Evolved Universal Terrestrial Radio Access (E-UTRA); Physical layer; Measurements”
 3GPP TR 36.777, “Enhanced LTE support for aerial vehicles,” 2017.
 Y. LeCun, Y. Bengio and G. Hinton, Deep learning. International Journal of Science Nature, vol. 521, pp 436-444, 2015.