AI Meets Network Requirements
Speaker: Marco Fiore
AbstractArtificial intelligence (AI) has permeated research in many scientific domains, and networking is no exception: machine learning is widely considered instrumental to the automation of network operation, and a vast related literature has emerged in recent years. Yet, networks do represent an exception when it comes to the unique requirements they impose to AI. Meeting these specifications is crucial to the viability and performance of AI-driven networking, and calls for a careful and dedicated design of machine learning models. In this talk, I will discuss challenges related to aspects like latency, computing or performance, and provide examples of how the integration of AI in network systems often requires advancing the state of the art in machine learning.
BioMarco Fiore is a Research Professor at IMDEA Networks Institute, where he leads the Networks Data Science group, and a co-founder and CTO at Net AI, a UK-based network intelligence company. He received MSc degrees from University of Illinois at Chicago and Politecnico of Torino, a PhD degree from Politecnico di Torino, and a Habilitation à Diriger des Recherches from Université de Lyon. Marco has held tenured positions at Institut National des Sciences Appliquées de Lyon and National Research Council of Italy, and has been a visiting researcher at Rice University, Universitat Politècnica de Catalunya, and University College London. Marco's research has received funding from the European Commission and national agencies in Spain, France and Italy, and has been awarded a number of prizes including the best paper award at IEEE INFOCOM. Marco is a former Marie Curie fellow and Royal Society visiting research fellow, a Senior Member of IEEE, and a Member of ACM. Marco's research interests are at the interface of mobile networks and data science.
Hands-on with a Battery-Free Robot Swarm
Speaker: Danny Hughes
AbstractSwarm robotics uses large collections of robots that work together in a cooperative yet decentralized manner to achieve their objectives, often drawing inspiration from biological systems such as insect colonies. However, power remains a critical problem for the swarm. Conventional batteries are slow to recharge, have limited lifespans and contain toxic chemicals which require special disposal techniques. This tutorial will explore a different future, by giving attendees hands-on experience with a novel battery-free swarm robotics platform that tackles these problems. CapBot combines supercapacitor charge storage with low power design to deliver a unique feature set, including: sub-cm accurate localization, full recharge in 16seconds, 51 min of autonomous operation at 0.74 km/h and Bluetooth networking. Each CapBot is also capable of performing rapid trophallaxis, where robots donate charge to each other in the field. Attendees will learn to program CapBots in C or Python, using standards-based tools to connect their robots to cloud based visualization and control software. Together, they will then use the system that they have built to perform an energy aware rescue mission using the CapBot.
BioDanny Hughes is an Associate Professor with the Department of Computer Science of KU Leuven (Belgium), where he is a member of the DistriNet research group and leads the Networked Embedded Software taskforce. Danny has a PhD from Lancaster University (UK) and has since worked in Brazil, China, and the US. His PhD focused on Peer-to-Peer (P2P) systems and his current research is on the Internet of Things (IoT) with a focus on ultra-low power and battery-free sensors and robotics. Danny has founded two spin-off companies, Isis Forensics (now Relative Insight, UK) and VersaSense (BE).
Introduction to Federated Learning: A Practical Guide for Researchers
Speaker: Charles Beauville
AbstractThis session will introduce the fundamentals of Federated Learning (FL), covering key use cases and exploring current challenges in the field. We will highlight open research problems and discuss how Flower AI, an open-source FL framework (https://flower.ai), can accelerate research and facilitate the transition from simulation to production. In the hands-on portion, participants will have the opportunity to join a live federation and send training tasks in order to develop a simple FL application. This should provide everything you need to start your FL journey!
BioCharles Beauville joined Flower AI Labs in its early days while completing my MSc at EPFL, and later conducted research on Asynchronous FL for his thesis at MIT’s CSAIL. Currently, he works as an ML Engineer at Flower AI Labs, contributing to the development and deployment of novel algorithms in production environments.
Towards Battery-Free Wireless Networks
Speaker: Marco Zimmerling
AbstractThe proliferation of cyber-physical systems (CPS), vital for industries and societal advancements, relies on wireless networks with untethered power sources, allowing embedded sensors to operate in remote or inaccessible locations. Batteries, with their limited lifespan and environmental impact, hinder scalability as replacing billions of batteries would be untenable and antithetical to the very sustainability efforts CPS are aiming to support. Battery-free technology promises zero-maintenance embedded sensing powered entirely by energy derived from renewable sources, such as solar and vibrations. Relying on volatile harvested energy, however, makes operation intermittent: the devices frequently lose power and need to recharge to resume operation. While it is known how to deal with these non-deterministic interruptions on a single device, we only begin to understand how to network multiple battery-free devices without relying on excessive external infrastructure like base stations. In this lecture, I will highlight the application requirements and underlying problems that make battery-free networking a formidable challenge. I will further review the scientific literature on the subject and describe the key principles behind a few selected solution approaches in more detail. I will conclude with an outlook on open questions and opportunities for future research.
BioMarco Zimmerling is a Full Professor at TU Darmstadt, where he has led the Networked Embedded Systems Lab since April 2022. Previously, he was a Full Professor at the University of Freiburg and, from 2015 until 2022, an Independent Research Group Leader at TU Dresden. In 2015, he completed his PhD in computer engineering at ETH Zurich in the group of Lothar Thiele. He holds a diploma degree in computer science from TU Dresden. For his diploma thesis project, he visited the groups of Thiemo Voigt and Per Gunningberg at RISE and Uppsala University. During his studies, he interned at IBM for over a year, including a six-month stay at the T.J. Watson Research Center. His research interests are in the area of cyber-physical systems, with a focus on wireless embedded systems. Overall, he aims to design and build real systems that are provably dependable, highly adaptive, and sustainable by design. His work has been recognized through several awards, including the 2022 ACM SIGBED Early Career Research Award, the 2022 SenSys Test-of-Time Award, and Best Paper Awards at EWSN 2022, ICCPS 2019, SenSys 2013, and IPSN 2011. More info at https://nes-lab.org/marco-zimmerling/
Build Your Own Private 5G Network with Software-Defined Radios
Speaker: Haitham Hassanieh
AbstractOne of the most critical trends in next-generation cellular networks is moving away from black-box proprietary hardware towards more open software-based RAN (Radio Access Network). Software allows us to better customize the cellular network to certain applications, slice the network, create virtual RANs that are run by different entities or provide different guarantees, and more easily test out new research ideas. Today, with open-source software, it has become easy and cheap for researchers to build and deploy their own private 5G networks. In this tutorial, we will learn how to build a 5G network using software-defined radios and open-source software like srsRAN for the radio access network and Open5GS for the 5G core. We will also learn how to do network slicing and resource allocation to independently customize the performance of each network slice.
BioHaitham Al Hassanieh is an associate professor in the school of Communication and Computer Science at EPFL. His research is in the areas of wireless networks, mobile systems, sensing, and algorithms. Before joining EPFL, he was a professor at University of Illinois at Urbana Champaign (UIUC) and he received his PhD from MIT in 2016. His PhD thesis on the Sparse Fourier Transform won the ACM Doctoral Dissertation Award, the Sprowls best thesis award at MIT, and TR10 Award for top ten breakthrough technologies in 2012. His research has received best paper awards at ACM SIGCOMM and ACM MobiSys. He is also the recipient of the NSF Career Award, the Google Faculty Research Award and the Alfred Sloan Foundation Fellowship.
Backscatter Communication: Principles and Impact of Parameter Choices
Speaker: Christian Rohner
AbstractThe tutorial offers an introduction to the principles and practical implementation of backscatter communication. Using the LoRea architecture as a case study, we will explore how various system parameters impact performance metrics such as range, data rate, and power consumption. Through a hands-on session, participants will experiment with parameter configurations and conduct their own performance analyses, gaining practical experience in designing and optimizing backscatter communication systems.
BioChristian Rohner is professor in computer systems at Uppsala University. His research interests include wireless communication, in particular low-power and battery-free system design. He has more than 8 years of hands-on experience with backscatter communication and is co-author of the LoRea backscatter architecture. The educational backscatter platform developed by the team has been awarded a best demo award at MobiSys 2023.
Benchmarking Low-Power Wireless Systems on Real-World Testbeds
Speaker: Carlo Alberto Boano
AbstractWhen performing research on low-power wireless systems, experimentation in real-world settings is often indispensable to validate the practical applicability and robustness of proposed solutions, as well as to convincingly prove their effectiveness to experts in the field and gain acceptance. Testbed facilities offer a convenient middle ground between simulation (which struggles to accurately reflect the complexity and dynamics of real-world scenarios) and full-scale real-world deployments (which come with logistical, financial, and scalability challenges); however, they do not inherently simplify experimentation (which remains complex, tedious, and time-consuming), nor do they ensure fair/objective comparisons and repeatable results. After highlighting the challenges and pitfalls of testbed experimentation, this lecture will emphasize the importance of rigorously benchmarking low-power wireless systems, introducing concrete methods and open-source tools developed by our group to simplify this process. These tools enable an autonomous execution of testbed experiments with a high degree of repeatability, provide a fair and impartial ranking of the performance of different solutions, and allow to reproducibly assess the robustness of a system under harsh environmental conditions.
BioCarlo Alberto Boano is an associate professor at Graz University of Technology, where he leads the Low-power Embedded Networked Systems (LENS) research group. He received a doctoral degree from Graz University of Technology in 2016 after obtaining a double Master degree from Politecnico di Torino and KTH Stockholm. During his post-doctoral tenure, he has been a visiting researcher at Uppsala University, the National University of Singapore, Tsinghua University, and the University of Trento. Carlo Alberto's research interests encompass the design of dependable and sustainable networked embedded systems, with strong focus on the efficiency and reliability of low-power wireless communications. His research has received several recognitions, including best paper/demo/poster/artifact awards at highly-reputed venues such as ACM SenSys, IPSN, SECON, EWSN, DCOSS, and WiMob. Carlo Alberto has participated as TPC member in more than 70 conferences/workshop in the area of networked embedded systems, and has recently served as general co-chair of EWSN'24, TPC co-chair of DCOSS'23, and as an editor for the IEEE Internet of Things Journal. More info at https://www.carloalbertoboano.com/.