Internship | AI-based management for sustainability in 6G Cell-Free Networks

About this position

Energy savings and sustainability are a big challenge in mobile networks, with the energy consumption continuously increasing due to the ever-increasing mobile network traffic.
Even though efforts are made to reduce the energy consumption of the mobile network in various ways, the radio infrastructure, in particular the base stations, are the components within the whole network to consume the largest amount of energy. A promising new architecture for achieving lower energy consumption while further enhancing the spatial performance fairness compared to traditional cellular networks, is cell-free networks (CFN).

CFNs consist of multiple simple low-power access points (APs), which are spatially distributed, and can be seen as a distributed MIMO system. Multiple of these APs can be assigned to serve a single device. The intelligence of the network (for example AP clustering, scheduling, and sleep mode management) is located in a centralized processing unit, which is coordinating and managing the APs. Because networks are designed to handle peak hour traffic, during non-peak hours some of the APs can go into ‘sleep mode’ in order to reduce the network energy consumption. Hence, sleep mode management (SMM) algorithms are developed to dynamically adapt the ‘mode’ (i.e. active or light/deep sleep) of the APs, according to the traffic load in the network and while still satisfying e.g. coverage and throughput performance requirements. The goal of this project is to deploy deep reinforcement learning (DRL) methods for the SMM of the APs. A realistic scenario will be considered for the evaluations, where APs are deployed on lampposts in the city centre of Amsterdam.

What will be your role?

During this project, a number of key activities will be performed:

• Gain deep understanding on cell-free networks, as well as DRL methods.
• Development of a DRL-based SMM algorithm and integration of the algorithm in an existing simulator of a realistic cell-free network. The developed algorithm will be evaluated in the simulator in different scenarios and configurations, including testing in different conditions/scenarios than in those trained.
• Implementation of state-of-the-art SMM algorithms from the literature, to be used as baselines for the evaluation of the developed DRL-based SMM algorithm.
• Derive key insights and conclusions with respect to the performance of the algorithm, as well as the challenges and gains of using DRL compared to heuristic algorithms.

This graduation project involves closely working with TNO employees. Further, the project will be conducted as part of the large nationally funded research programme ‘Future Network Services (FNS)’, bringing together more than 60 partners for a targeted period till the middle of 2030. The proposed project is envisioned to involve collaborations and/or regular interactions with other FNS partners.