Motivation

Climate change is the most important challenge of our century. Reducing and reverse its destructive effects is top priority for all countries, especially for the research community.

In the other hand, even tough the transition to more greener sources of energy is progressing, there still societies that rely strongly on fossil fuels (especially the areas with largest population growth). Reducing the power consumption of systems by the use of more energy efficient designs is one way to address this global problem.

As the data traffic continue to grow exponentially, mobile networks have been expanded and radio networks have become a relevant contributor to power consumption globally. While in 2015 the contribution was of 1.15%, and contributed to 0.53% of the global carbon emissions*, it is expected to grow as so the number of new devices and use cases expected in 5G and beyond. Designing more efficient telecommunication systems has a direct impact on carbon emissions and therefore on climate change.

My area of research is focused on making the radio access network more energy efficient, especially for beyond 5G and 6G systems. I consider two main sources of power consumption in such a systems: radiated power and power consumed in the electronics during data processing and data movement. Radiated power can be reduced by increasing the number of antennas, which allows to focus the radiated power more accurately (therefore no radiating in unintended locations). However, an increment of number of antennas may translate to an increment of data to process and move within the system. I conduct research in system level architecture, exploring different topologies, and distributed processing algorithms that not only have reduced computational complexity but also has low requirements in terms of data movement.

Research

My research topic is efficient algorithm-architecture co-design for large-antenna array systems (Massive MIMO and Large Intelligent Surfaces). It covers:

• Identification of potential hardware limitations and bottlenecks in centralized architectures.
• Exploration of different solutions at architecture and algorithm level to alleviate these issues.
• Proposal of different decentralized solutions with distributed processing, including analysis and trade-offs between different system factors: communication performance, interconnection data-rate, computational complexity, latency, memory, etc.
• Exploration of machine learning methods to support wireless-based positioning.

For a detailed list of publications please visit my scholar page.

Projects

Massive MIMO

Massive MIMO has moved from research topic to physical product under current deployment for 5G networks in a decade. In despite of numerous advantages of this technology there are still implementation challenges that need to be solved in order to make these systems more energy-efficient. My contributions to this project ranges from proposing different system topologies to architecture-algorithm co-design to ensure a more energy efficient and easy to implement system. See [TSP2020] for mroe details.

Large Intelligent Surfaces (LIS)

While Massive MIMO is considered a step forward in terms of spectral efficiency and beamforming capabilities, further steps are needed in that direction to accommodate for the expected traffic growth and number of devices in near future. Large Intelligent Surfaces go in that direction, by placing hundreds or thousands of antennas in a certain scenario relatively close to users. While the theoretical benefits are very promising, there are many implementation challenges to be addressed, which are not covered in current literature. My contribution to this project aims to fill this gap. For more details please see [submitted-to-journal].

Positioning

In 5G, new use cases will be relying on accurate positioning. While GNSS based-systems work well in outdoor areas, is not the case in indoor scenarios. There is a need to determine user location by means of the radio signals transmitted to the base station from the user during normal operation. In this project we leverage machine learning techniques to acquire user location by use of channel state information. In this work, we present a positioning algorithm suitable for distributed processing in a Large Intelligent Surface, where instead of point estimate, we provide a probabilistic description of the user location. Different parts of the surface provide a probabilistic estimate that can be fused with other estimates providing sub-wavelength accuracy in our simulations. For more details refer to [Globecom2021]