The PhD thesis will take place at INSA Rennes, IETR Laboratory. It is fully funded and is expected to start in Autumn 2024.
Context
As we embark on the journey towards 6G wireless networks, a multitude of hurdles must be overcome by engineers and researchers. The imperatives of high throughput, low latency, and ultra-reliable communications loom large on the horizon. Concurrently, the escalating need for eco-conscious practices in communications is increasingly evident, stemming from the necessity to reduce the ecological footprint of mobile radio networks. Enhancing the energy efficiency of wireless networks is of paramount significance, particularly in light of forecasts suggesting that communications could contribute up to 14% of global CO2 emissions by 2040.
In this dynamic and complex network ecosystem, revolutionary technologies are essential to effectively meet the diverse requirements imposed by technical, environmental and societal concerns. To this end, researchers advocate three main factors enabling more efficient and greener communications: (i) massive distributed antenna systems, also known as cell-free massive multiple-input-multiple-output (CF-mMIMO) networks, (ii) energy-efficient resource management techniques and (iii) signal processing solutions assisted by artificial intelligence (AI).
CF-mMIMO technology promises significant improvements in mobile networks by boosting macro-diversity and ensuring consistent spectral efficiency (SE) across coverage areas. It also sidesteps the issue of intercell interference, a common problem in today's cellular networks. This marks a paradigm shift from conventional systems, overcoming their limitations and meeting the challenges of B5G and 6G wireless networks. On the other hand, AI tools offer valuable support by devising efficient solutions to complex optimization problems, thus overcoming issues related to high complexity and latency.
Nevertheless, efficient spectral and power management techniques rely on the availability of high-dimensional channel-state information (CSI) gathered at numerous multi-antenna Access Points (APs) over broad bandwidths. Consequently, a substantial volume of CSI must be processed and exchanged across networks, and at fast rates, especially when dealing with high user mobility. To address this challenge, the framework of channel charting (CC) has been introduced to leverage the spatial information inherent in CSI. This involves mapping the high-dimensional CSI into a lower-dimensional chart, where the relative positions of users are preserved. As a result, CC can be viewed as a method for compressing CSI. To achieve this, CSI is first utilized to extract channel features, which are subsequently employed to learn a charting function in an unsupervised fashion.
Objectives and methodology
The aim of this thesis is to introduce innovative approaches to improve B5G and 6G communications, focusing on the maximization of the network energy efficiency. This entails ensuring the optimal balance between the ecological footprint and spectral efficiency. Effective interference management holds paramount importance in attaining these objectives. Hence, the research will focus on optimizing resource allocation while using non-orthogonal multiple access (NOMA). NOMA involves assigning two or more users to the same spectral/temporal resource through suitable power multiplexing, thereby enhancing the system SE. Therefore, the study will consider a joint optimization of NOMA user clustering, MIMO precoding, power allocation, and AP selection.Instead of relying on complete CSI, channel charts will be employed to accomplish these tasks, thereby maintaining a moderate level of system complexity and signaling exchange.
The candidate will start by conducting a comprehensive review of the state-of-the-art in CC, resource allocation techniques and CF-mMIMO networks. Subsequently, they will focus on developing and implementing novel strategies for efficiently managing spectral, spatial, and energy resources. Special emphasis will be placed on identifying the most appropriate cell-free network configurations (centralized, distributed, hybrid) to fully leverage the advantages of channel charting.
More specifically, the work will start by identifying the most suitable channel distance measure to the resource allocation task. This implies determining appropriate channel features at a first step, depending on the application and context at hand.
In a second phase, dimensionality reduction will be performed, based on the selected features. To accomplish this, various machine learning methods can be envisaged such as contrastive learning (Siamese or triplet deep networks), manifold learning (Isomap, Multi-Dimensional scaling), autoencoders, among others. Particular attention will be devoted to identifying the most suitable algorithms for the distributed scenario inherent in CF-mMIMO systems.
In order to guarantee the practicality of the proposed methods in realistic scenarios with user mobility and environmental changes, the temporal evolution of the channel charts will be considered by relying on online learning techniques that perform an adaptation of previously generated charts. These charts are then continuously used to optimize resource allocation.
Keywords
Cell-free massive MIMO networks, channel charting, non-orthogonal multiple access, resource allocation, machine learning.
References
[1]O. T. Demir, E. Bjornson, and L. Sanguinetti, “Foundations of User-Centric Cell-Free Massive MIMO,” Foundations and Trends® in Signal Processing, vol. 14, no. 3-4, pp. 162–472, 2021.
[2]X. Chen et al. "User pairing and pair scheduling in massive MIMO-NOMA systems", IEEE Communications Letters, vol.22 no.4, pp. 788-791, 2017.
[3]C. Studer, S. Medjkouh, E. Gonultaş, T. Goldstein and O. Tirkkonen, "Channel Charting: Locating Users Within the Radio Environment Using Channel State Information," in IEEE Access, vol. 6, pp. 47682-47698, 2018.
[4]L. Le Magoarou, T. Yassine, S. Paquelet and M. Crussière, "Channel charting based beamforming," 2022 56th Asilomar Conference on Signals, Systems, and Computers, Pacific Grove, CA, USA, 2022, pp. 1185-1189.
[5]L. Ribeiro, M. Leinonen, H. Djelouat and M. Juntti, "Channel Charting for Pilot Reuse in mMTC with Spatially Correlated MIMO Channels," 2020 IEEE Globecom Workshops, Taipei, Taiwan, 2020, pp. 1-6.
Candidate profile
We are seeking candidates with an engineering degree and/or Master's degree in Telecommunications or a directly related field. The candidate should possess a strong background in Digital communications, Mobile networks, Signal processing, Machine learning and proficiency in programming languages such as Matlab and Python.
Practical information
Location: The thesis will be take place at the IETR laboratory, INSA, Campus of Beaulieu, Rennes.
Starting date: September or October 2024.
Duration: 3 years.
Salary: The gross salary starts at 2100 Euros per month.
To apply
Please send a CV (including at least two references with their contact information), a motivation letter, copies of all academic records and grades (preferably with rankings), and optionally, 1 or 2 recommendation letters to the thesis supervisors:
Joumana Farah:joumana.farah@insa-rennes.fr
Matthieu Crussière:matthieu.crussiere@insa-rennes.fr
Luc Le Magoarou: luc.le-magoarou@insa-rennes.fr
Only complete applications will be considered. All documents must be in either French or English.
(c) GdR IASIS - CNRS - 2024.