Wireless Networks and IoE

Proposta de tesi

Investigadors/es

Grup de recerca

Future-compatible industrial communication technologies

The digitalization of the industry is a step further to achieve a digital society. Such digitalization will enable more efficient industries with major impact in the quality of work and the generated industrial value and competitiveness. To achieve such digitalization, massive connectivity will be progressively introduced to industrial processes, mainly by extending existing machinery interfaces and integrating to existing industrial infrastructures and information systems on a first stage. This integration of Information Technologies (IT) and Operational Technologies (OT) by itself imposes challenges beyond what is envisioned by the 5G architectures since, industrial processes are not only critical in terms of reliability, latency and security but also in their architecture, heterogeneity and ownership model which will require more flexible network architectures to complement those envisioned by 5G.  In this research proposal we aim to address industrial requirements to support digitalization. The research work will be centered in the study and development of mechanisms that enable industrial communication interfaces to be future and backward compatible by exploring novel mechanism and capabilities to self-upgrade, exploiting software-based radio interfaces and enabling any new protocol on any frequency band to be executed when needed. Supporting that features will permit in the future that old machinery is connected to novel systems in an efficient manner, supporting better evolution and not limiting industrial vendors to stick to legacy technologies.

 

Dr Xavi Vilajosana

WINE

RFID-based ubiquitous sensing

Automated ubiquitous sensing would represent a major change in current societal challenges in terms of efficiency (i.e. automatization), sustainability (better models and decision-making), and social wellbeing (security and improved job market). Automated ubiquitous sensing would, for instance, detected humidity or thin water ponds on a road avoiding possible car accidents in a smart city context, or detect people-object interactions in an ambient assisted living environment.

Current solutions for automated sensing, including active and passive sensors, use specific electronics for sensing besides wireless communications technologies to transmit the measurements. This approach implies battery management complexity and/or expensive customized technology, being inappropriate for actual ubiquitous sensing.

This proposal aims to exploit passive low-cost communication technologies like Ultra High Frequency (UHF) Radio Frequency Identification (RFID) to act as ubiquitous sensors, by means of analyzing context effects on the RFID labels performance (i.e. see [1,2,3]). The goal is to perform the first steps towards a complete environment digitization thanks to actual ubiquitous sensing.

[1] Melià-Seguí et al (2014): http://dx.doi.org/10.1109/IOT.2014.7030112

[2] Hasan et al (2015): https://doi.org/10.1109/RFID-TA.2015.7379812

[3] Le Breton et al (2017): https://doi.org/10.1109/RFID.2017.7945594

Dr Joan Melià

Dr Xavi Vilajosana

WINE

Context-aware applications for ambient intelligence

Smart cities are a new scenario where a variety of data sources open the door to innovative applications, with the final goal of improving the quality of life of citizens, industry competitiveness and government. The Internet of everything is responsible for collecting and transmitting this data at different levels, by using a wide variety of sensors (including peoples' smartphones). We propose to make use of contextual intelligence tools and techniques (like data mining or statistical learning) to process signals (accelerometers, acoustic, etc.) with the overall goal of transforming the above collected data into useful information, enabling new applications or improving existing ones within the ambient intelligence context.

Dr Joan Melià

Dr Carlos Monzo

WINE

Ultra-reliable low latency industrial communication technologies

The digitalization of the industry is a step further to achieve a digital society. Such digitalization will enable more efficient industries with major impact in the quality of work and the generated industrial value and competitiveness. To achieve such digitalization, massive connectivity will be progressively introduced to industrial processes, mainly by extending existing machinery interfaces and integrating to existing industrial infrastructures and information systems on a first stage. This integration of Information Technologies (IT) and Operational Technologies (OT) by itself imposes challenges beyond what is envisioned by the 5G architectures since, industrial processes are not only critical in terms of reliability, latency and security but also in their architecture, heterogeneity and ownership model which will require more flexible network architectures to complement those envisioned by 5G.  In this research proposal we aim to address industrial requirements to support digitalization. The research work will be centered in the study and development of mechanisms to enable ultra-reliable and low latency industrial wireless communications. The envisioned communication technology should support robotics eliminating the need of wires, while ensuring high reliability (99.999%), low latency (<1ms) and secure links. The research work will explore the features provided by novel physical layer technologies, based on mmWave bands and exploit redundancy mechanisms to achieve the desired performance.

Dr Xavi Vilajosana

 

Dr Ferran Adelantado

WINE

Data driven 5G networks: an Artificial Intelligence approach

The continuous increase in the mobile data traffic demand has posed significant challenges in the design of 5G networks. Although video traffic has dominated the surge in mobile data traffic demand so far, the growing importance of new services with diverse Quality of Service (QoS) requirements stretches the networks to their capacity. In that sense, three service categories have been identified: enhanced Mobile Broadband (eMBB) services, Ultra-Reliable Low-Latency Communications (URLLC) service and massive Machine-Type Communications (mMTC) services.

The ability to serve such diverse traffic will be determined by the flexibility and adaptability of the network. That is, tailored virtual networks will have to be built on top of the physical network to meet the requirements of each specific service. In this context Artificial Intelligence (AI) will play a key role. The exploitation of the data available in future SDN-based networks will be the enabler to boost the spectral efficiency and the design of a set of new AI-based Radio Resources Management (RRM) algorithms.

Artificial Intelligence has been extensively studied and applied in scientific fields such as Computer Vision. However, its application to communications networks is still in its infancy. This PhD thesis aims to analyse existing Machine Learning and Deep Learning and evaluate the feasibility and advisability of its application in 5G mobile networks.

We are looking for highly motivated, enthusiastic junior scientists, with an MSc in electrical engineering or related fields, aiming at significantly improving their career. Excellent research skills and analytical abilities are required, fluency in English (spoken and written), proactive communication skills and problem solving as part of a team, strong record keeping, great work ethic and initiatives are essential characteristics.

Requirements:

Candidates must have a strong mathematical background, fluency in English (spoken and written) and excellent analytical and writing skills. Experience in artificial intelligence and networks (specially wireless networks) will be highly appreciated.

Related references:

[1] R. Li et al., "Intelligent 5G: When Cellular Networks Meet Artificial Intelligence," in IEEE Wireless Communications, vol. 24, no. 5, pp. 175-183, October 2017. doi: 10.1109/MWC.2017.1600304WC

Dr Ferran Adelantado WINE

Wireless network optimization via experience share

Wireless networks are facing an exponential increase in traffic demand along with a massive increase in density of devices and heterogeneity. Given such complexities, traditional optimization frameworks based on analytical characterization of the network behaviour are expected to be intractable or provide answers difficult to understand and implement in practice. In contrast, learning techniques (such as reinforcement learning) can inherently capture the resulting complex interactions among devices and provide optimality by learning from experience. In this thesis, the use of reinforcement learning mechanisms will be studied in order to overcome the limitations of analytical characterization with the goal of achieving efficient and fair spectrum usage in complex wireless systems, such as 5G networks and in the IoT landscape. The main challenges in applying reinforcement learning to wireless networking optimization are: a) to reduce complexity of network abstractions, b) to amortize the cost of exploratory tasks used to obtain new knowledge, and c) to reduce convergence times. This thesis will explore the use of experience share among multiple network devices to address these challenges and make reinforcement learning applicable to wireless network optimization.

Dr Cristina Cano

Dr Xavi Vilajosana

WINE