Safehaul: Risk-averse learning for reliable mmWave self-backhauling in 6G networks:This paper is accepted for

presentation and further publication in the upcoming IEEE Conference on Computer Communications (INFOCOM

2023). In the paper, a millimeter wave backhauling scenario is considered and a risk-averse learning-based solution

for link scheduling is proposed. We define risk as the probability that the latency in a backhaul link is larger than a

safety threshold. In contrast to the state-of-the-art solutions, which focus on average performance optimization, we

ensure reliability by jointly optimizing the average performance and minimizing the losses in the tail of the

performance distribution. Moreover, we extended NVIDIA’s newly released GPU-accelerated simulator Sionna. This

work provides a good foundation for the project ASUCAR because considering reliability is a promising first step

towards achieving sustainable and resilient radio access networks.

Deep reinforcement learning for task allocation in energy harvesting mobile crowdsensing: This paper was

recently presented at the IEEE Global Communications Conference (Globecom 2022). It considers a crowd-sensing

scenario in which a mobile crowd-sensing platform allocates sensing tasks to its associated sensors. Furthermore,

sensors capable of harvesting energy are considered to ensure continuity in the energy supply. To make optimal

allocation decisions, the mobile crowd-sensing platform requires perfect non-causal knowledge about the dynamics

of the system, e.g., channel coefficients of the wireless links to the sensors or the amounts of harvested energy.

However, in practical scenarios this non-causal knowledge is not available. Our proposed approach overcomes this

problem by exploiting reinforcement learning. Specifically, we propose a novel Deep-Q-Network solution to find the

task allocation strategy that maximizes the number of completed tasks in dynamic environments. By exploiting

energy harvesting and maximizing the system’s lifetime, this work is part of my current research effort towards

sustainable wireless networks.

Resource Allocation in mMTC Networks: A Statistical Physics Approach: This work is currently on the last

preparation round before its submission. It addresses the problem of allocating time-frequency resources in a

massive machine type communication scenario with the goal of minimizing the interference in the system. As the

problem is known to be NP-hard, it cannot be solved optimally in polynomial time when large networks are

considered. To overcome this challenge, we propose the use of the Survey Propagation method from statistical

physics. Survey propagation is a message-passing algorithm to solve constraint satisfaction problems and

traditionally used in spin glasses. This work is the starting point for the design of scalable centralized resource

management methods. The results obtained indicate that for networks that have a tree-like structure, survey

propagation is able to outperform state-of-the-art approaches.