Novel Scenarios for the Wireless Internet of Things

The seminar commenced with a keynote speech by Dr. Victor Bahl, a Technical Fellow at Microsoft. This was followed by a pair of speak-and-spark sessions, during which each attendee delivered a brief presentation on their research, facilitating mutual familiarization of everyone’s areas of interest. The inaugural day culminated with an invited talk given by Dr. Olga Saukh. Moving to the subsequent day, the program featured four invited talks along with two break-out sessions. During these break-out sessions, attendees were organized into smaller groups, each focusing on a specific IoT research theme. The final day was dedicated to wrapping up the event and engaging in discussions. Leaders of each break-out group presented summary reports of their respective discussions, effectively concluding the seminar. The detailed agenda is listed in Table 1.

The seminar centered around six comprehensive research topics within the realm of the Internet of Things (IoT). These topics encompassed a diverse spectrum, including underwater communication, implantable IoT devices, airborne communication systems, integrated communication and sensing technologies, the intersection of artificial intelligence and IoT, and the optimization of energy efficiency in IoT applications. Through in-depth presentations and break-out discussions on these subjects, participants gained valuable insights into the cutting-edge advancements, challenges, and opportunities shaping the future landscape of IoT technology. The seminar fostered a collaborative environment where participants exchanged ideas and discussed potential research collaborations. Overall, the seminar was considered a success, building the next cornerstone for the wireless Internet of Things.

Furthermore, the extensive research discussions held during the break-out sessions have ignited a multitude of captivating concepts for the upcoming wave of IoT applications. This enthusiasm culminated in the creation of four technical reports, each focusing on distinct research realms:

• $\mathrm{■}$

  Energy Efficiency of IoT (§4.1)

• $\mathrm{■}$

  Integrated Communication and Sensing (§4.2)

• $\mathrm{■}$

  Advancements in Medical IoT (§4.3)

• $\mathrm{■}$

  Airborne Internet of Things (§4.4)

• $\mathrm{■}$

  Impact of AI on IoT (§4.5)

These reports delve into the challenges faced and chart potential trajectories for future progress in these specified areas.

In this talk, I will delve into the endeavors of our research team aimed at enhancing mmWave self-backhauling. Our focus has been on addressing challenges related to beamforming and route selection within self-backhauled networks. We’ve explored solutions using reinforcement learning as well as traditional optimization methods. Additionally, I will elaborate on our strategies for countering adversarial WiFi sensing and mitigating user tracking through the application of radio fingerprinting techniques. Furthermore, I will provide a concise overview of our advancements in Reflective Intelligent Surfaces (RISs) development, encompassing both sub-6GHz implementations utilizing RF-switches and 60GHz designs employing Liquid Crystals.

The Internet of Things (IoT) is a term generally used for a (too) wide variety of hardware, platforms and protocols. In this talk we focus less on cloud/edge IoT aspects and hardware based on CPU/GPU, and more on things operated “beyond” edge computing: we focus on ultra-low power hardware, software and protocols running on microcontrollers. This is a category of devices the IoT relies heavily upon. To be more concrete, RFC7228 distinguishes several classes of such devices, for example Class 1 gathers devices providing memory budgets of 10 kB RAM and 100 kB flash. With such a target in mind, we quickly overview the key aspects of relevant network protocols (distributed algorithms, standard specifications, interoperable implementations) and we flip through the list of relevant working groups that are currently active at the Internet Engineering Task Force (IETF/IRTF) standardizing IP protocol adaptations for ultra-low power IoT. In parallel, we overview the embedded software platform developments that take place, in order to provide open source interoperable implementations of these open standards. We overview developments based on RIOT, a small-footprint operating system that runs on a wide variety of microcontroller boards and aggregates various libraries and network stacks.

We are at the beginning of an unprecedented opportunity for information technology startups and the established cloud industry to become a part of the next generation telecommunications infrastructure. Together, we can radically change it through softwarization, AI and edge computing. I will describe the scientific advances and business needs for bringing us to where we are today and then cast an eye to the future in sharing with the audience a vision for where things are going with telecommunications, including key enablers and potential surprises on the horizon. This will set the context for describing the opportunity ahead for the engineering and research community in the next several years, and beyond, as we stay at the forefront of the modernization that will enable ubiquitous computing via telecommunication networks propelled by innovations in AI and edge computing. I will next move into near-term strategy with Microsoft standing up of a new business division called Azure for Operators (AFO). I will describe the vision that led to the creation of AFO, its mission, and products, and the significant technical and scientific challenges, which we are pursuing. When overcome they will lead to the inevitable convergence of two massive industries that will open up a new decade of opportunities for universities, research institutes, established companies, and startups.

Future wireless communication systems will be increasingly software-defined, with AI-native components to optimize performance. With current Software Defined Radios, there are powerful general-purpose hardware platforms that could drive such systems. What we are missing are software frameworks that can make efficient use of the heterogeneous compute resources available on those platforms. In this talk, I will briefly discuss my research effort on filling this gap by designing and implementing novel real-time signal processing systems that provide full control of the data flow through custom schedulers and native support for hardware accelerators like FPGAs and GPUs. This can serve as the enabler for a new generation of wireless networks that use AI not just for multi-objective optimizations but as the base for a new system design that goes beyond traditional protocol structures and complex error handling to create more robust and more resilient systems.

Access to even basic medical resources is greatly influenced by factors like an individual’s birth country and zip code. In this talk, I will present my work on designing intelligent mobile systems for equitable healthcare. I will showcase three systems that are not only interesting from a computational standpoint but are also having real-world medical impact. The first system can detect ear infections using only a smartphone and a paper cone. The second system enables low-cost newborn hearing screening using inexpensive earphones. Lastly, I will present an ambient sensing system that employs smart devices to detect emergent and life-threatening medical events such as cardiac arrest. Through these examples, I will demonstrate how new computational and sensing techniques that generalize across hardware and work in real-world environments can help to address pressing societal problems.

As the number of connected devices increases, the stress on the wireless bandwidth available keeps increasing. Researchers across the area of wireless systems are building novel communication and sensing systems to overcome these limitations by addressing deployment domain-specific problems such as, Space-Based Internet, Low-Power Low-Bandwidth IoT Sensors for Agriculture, and Low-cost Sensing Solutions for the supply chain. Yet, many of these research prototypes are considered as research artifacts which provide basic connectivity and sensing, yet lacking much of the improved utility (as typically mentioned in many research contributions).

This talk will focus on the important wireless problems at the frontier of these domains to improve end-user acceptance and utilization of these impressive technologies. This talk will present several important questions that a typical user faces today and how academics and industry researchers can collaborate on these problems to alleviate these issues. Finally, this talk will provide the base for interesting discussion throughout the day of meetings with researchers on brainstorming these ideas and building collaborations.

I will explore the integration of Micro-Electro-Mechanical-Systems (MEMS) filters with innovative sparse recovery algorithms, enabling energy-efficient IoT devices to effectively capture wideband spectra. I will showcase the practical implementation of this approach in two distinct applications. Firstly, I will introduce an energy-conscious spectrum sensing approach that empowers low-power IoT devices to detect wideband channel occupancies. This capability enables these devices to opportunistically utilize unoccupied channels, facilitating dynamic spectrum sharing. Secondly, I will unveil a precise self-localization technique for IoT devices. This method harnesses existing ambient 5G communication signals without necessitating any alterations to the 5G infrastructure. This innovation offers accurate self-localization capabilities without requiring modifications to the underlying infrastructure.

We aim to develop efficient mobile networked systems that include machines, vehicles, and humans. Hereby we propose novel architectures of such networked systems, analyze the performance, and experimentally evaluate the overall capabilities of the wireless network and often context-sensitive algorithms for the wireless link or overall networked system, including predictive approaches. Two recent application scenarios for which we set-up small wireless networked testbeds to work on novel solutions target systems for human-drone teams in indoor scenarios, where the drone’s camera is the primary sensor, used to navigate and interact with the human and the environment. We find that a distributed control architecture leveraging direct Wi-Fi links is feasible and beneficial for such systems having an AI or machine learning component, yet common Wi-Fi cannot provide time guarantees which would be necessary for many real-time applications. However, this architecture enables the successful deployment of sophisticated applications such as drone learning from human emotions, which is one of our current research projects. In this talk, I will briefly introduce some of our current research results on these research threads.

In this talk, I will present three distinct advancements in the realm of sensing and interaction technologies. Firstly, I will introduce an innovative approach to inaudible attacks on smart earbuds, overcoming limitations of existing methods by leveraging both direct and reflective paths for attacking signals, thereby enhancing signal-to-noise ratios. Secondly, I will present a 3D Hand Posture Reconstruction system utilizing 2D Rolling Fingertips, which capitalizes on active optical labeling and exploits rolling shutter effects in smartphone cameras to achieve improved 3D tracking for complex scenarios like underwater environments and specialized applications like virtual writing. Lastly, I will present a low-power system for precise face touch detection, integrating wrist inertial and novel finger vibration sensors with a cascading classification model for efficient gesture filtering. This system achieves a high F-1 score of 93.5% for detecting face touch events, while maintaining minimal power consumption, making it suitable for prolonged usage with battery-powered devices. Each advancement is supported by practical prototypes and evaluations, showcasing their potential across a range of real-world scenarios.

Battery-less devices revolutionize conventional battery-dependent setups by integrating ambient energy harvesting technologies, thereby enabling a sustainable Internet of Things (IoT) paradigm with minimal maintenance requirements. However, the intermittent nature of ambient energy sources presents challenges. These devices frequently encounter energy shortages, causing intermittent computation where active phases alternate with inactive ones due to waiting for adequate energy. Given their limited energy reservoir, optimizing battery-less device operation is crucial to maximize computational output from ambient energy. This presentation introduces our research effort addressing these concerns. We begin with the introduction of ALFRED, a virtual memory abstraction and compilation pipeline tailored for mixed-volatile platforms. ALFRED autonomously identifies optimal program state mappings between volatile and non-volatile memory, enhancing efficiency. Our focus also extends to ensuring consistent efficiency in battery-less devices. We’ve developed a system design to efficiently regulate supply voltage and clock frequency, even in hardware-limited devices lacking dynamic voltage and frequency scaling support. Implementing two hardware/software co-designs capturing these aspects, our approaches reduce battery-less device energy consumption by up to 170% and significantly decrease workload completion time, enhancing their overall performance. Through these advancements, we contribute to the realization of robust and optimized IoT systems powered by ambient energy sources.

The Internet of Things (IoT) is making a great impact on our lives and changing the way we interact with others, the environment, and even machines. Now there are two considerations for IoT space. (a) The IoT devices deployed on the ground need a reliable connectivity backbone. (b) Space technology is seeing unprecedented growth since space subsystems are becoming smarter, more reliable wireless links, and further miniaturized – all these without the need for radiation-hardened components in Low Earth Orbit (LEO). The above two aspects lead us to a new domain of IoT called the Space Internet of Things (Space-IoT). In recent times, there has been a huge interest and investment in Space-IoT. Space can be a suitable platform to solve many of the current and future problems in IoT, and the possible solutions are yet to be explored in depth in the space environment. In this talk, I will discuss my group’s research effort in this research domain.

The aquatic world, with more than 70% of the Earth covered by water, plays a critical role in our society and the economy, as recognized by many organizations including the United Nations and the World Wildlife Fund, which estimated the economy value related to the aquatic world – also called Blue Economy – to be at least US$24 trillion. Yet, currently, more than 80% of the ocean is unmapped. This talk delves into the challenges and opportunities of deploying an Internet of Underwater Robots and Sensors (or more generally, Things) to explore the underwater world and support the Blue Economy, particularly focusing on underwater sensing and communication. Supported by lessons learned from actual field experiments, I will present first the general requirements for such underwater robots and sensors to operate in such environments; second, the technologies used in the current deployments, highlighting gaps and open problems; and finally, current research that shows research opportunities in the space of Ocean Internet of Things to achieve the long-term goal of ocean IoT, to support large scale aquatic applications, such as environmental monitoring or archaeological exploration.

AI is utilized in various IoT applications, ranging from environmental monitoring to home automation and production processes. However, the most concerning applications of AI are in safety-critical systems, such as transportation, medicine, and control, because incorrect use of AI can have devastating consequences. A significant amount of current research effort is focused on increasing the robustness of deep models or implementing safe and low-resource post-deployment domain adaptation techniques to manage domain shifts. In this talk, I will provide a concise overview of the state-of-the-art in the field and present the latest research results from my team. We discover essential properties of the loss landscape geometry and leverage these to improve AI safety across a broad range of applications, particularly in scenarios where resource constraints are at stake.

Large constellations of Low Earth Orbit satellites promise to provide near real-time high-resolution Earth imagery. Yet, getting this large amount of data back to Earth is challenging because of their low orbits and fast motion through space. Centralized architectures with few multi-million dollar ground stations incur large hour- level data download latency and are hard to scale. We propose a geographically distributed ground station design that uses low-cost commodity hardware to offer low latency robust downlink. We also discuss intersections of compute and networks in this emerging context. We believe our work engineers a paradigm shift similar to the one from super-computers to data centers, where software architectures can extract high performance from commodity hardware.

Mobility is the essential norm in human society. As today’s wireless and mobile networking technology already brings 99% usable connectivity between our personal devices and (edge) cloud services, the remaining fragmented 1% connectivity scenarios, including but not limited to extremely lower power, high mobility, and massive access, which still face domain-specific networking challenges. More importantly, these minority cases are likely to turn over into the majority tomorrow as our global society is consciously evolving in the direction of energy and production efficiency improvement. This talk will introduce our recent efforts towards the vision of “Ubiquitous Mobile Connectivity”. Specifically, we will present the design, implementation and deployment experience of our mobile RFID, VILD and multipath networking system for improving scalability, availability and reliability in logistics, Vehicular-to-X and high-speed railway networks.

The proportion of older adults (aged 65 and older) worldwide has been increasing steadily over the past 40 years. Among older adults, mobility is a crucial indicator of functional status, and a predictor of quality of life and longevity; hence, it is often called the sixth vital sign. Mobility encompasses not only the physical activities of older adults, and the performance of specific maneuvers such as sit-to-stand, walking or climbing stairs, but also participation in society (e.g., the ability to drive, accessibility to public transportation). Recent advancements in mobile technologies, artificial intelligence, embedded and sensing devices create exciting opportunities to address mobility challenges faced by the aging population. However, the shortage of quality, labeled, free-living data from target populations remains to be a main barrier in developing and applying effective solutions for continuous monitoring, analysis and assessment.

My research work aims to mitigate the data scarcity problem in characterizing old adult activities and answer “what”, “when” and “how well” questions regarding their functional mobility. We take a multi-pronged approach. Directly addressing data availability, we developed ChromoSim, a cross-mobility IMU data synthesizer, CROMOSim, that can transform abundant data from vision, motion capture systems to IMU data. At the modelling level, we devised efficient invariant feature learning framework that can easily adapt to novel subjects or devices and can handle noisy labeled data collected in the wild.

We, humans, already use 50% more energy moving information than moving airplanes around the world. Communication is central to our societies but it is taking a toll on the earth. We want to use a free, abundant, and natural resource for wireless communication: sunlight. Similar to the way you can use a mirror to communicate by reflecting light, our aim is to exploit optical devices that can change their reflective and transmissive properties to send information, but without you noticing any flicker. In this manner, objects will be able to talk to each other using daylight, an eco-friendly solution. In this talk, I will present some of our recent work in this area and its potential applications.

Internet-of-things is also unlocking the potential to augment devices not typically deployed statically as a sensor but instead a mobile device that can move around and sense different environments using the same sensor. One such direction of research exploration is understanding and augmenting the opportunity of airborne internet-of-things where devices with batteries/solar panels spend several seconds (insects), minutes (UAVs/drones/blimps), hours (airplanes) and even days (satellites) away from power.

There are tremendous opportunities that can be enabled by effectively capitalizing on the ability of these sensors to move around and provide connectivity. For instance, satellites and drones can act as data grabbers from terrestrial sensors with poor connectivity. Satellites and aircrafts can provide ocean coverage for various capabilities (localization, tracking, SOS, communication) that we traditionally take advantage of in well-connected environments. There has also been demonstration of using satellites for cellular backhauls for tracking global shipments by Telefonica. Further, there are new opportunities in exploring new capabilities such as edge computing and mesh networking in space. On the UAV side, enabling better autonomy for coordination of drone swarms for various sensing, transportation and entertainment applications remain important. Finally, planes can act as an intermediate zone of innovation between the drones which are battery constrained and the satellites which are communication constrained to assist the other two in enabling the same applications. It is also critical to not just build these services but also specializing them for commercial and environmental applications in real-world industries, such as mining, oil and gas, and agriculture. There are also upcoming deployments that cannot be classified into the above three categories but can enable new applications such as Insect IoT and Balloons IoT.

These opportunities can only be enabled by solving real world system challenges across the cyber-physical system stack from embedded and communication systems to better software applications that maximize the utilization of the limited resources at the sensors. One such challenge for satellites that is gaining increased exposure is the problem of reliability provided by satellite constellation-based internet. Indeed, a recent story of deployment in Ukraine highlighted this issue at a recent tutorial at MobiCom. Further, coordination of satellite communication directly to the sensors on the ground remains an open problem. Another important problem for satellite deployment is sustainability of deployment and debris reduction of existing satellites. In fact, recent efforts have taken place to use the 4th stage rockets lost in the orbit for satellite applications. Further, it was also described that most LEO satellite deployments deplane in orbit and get melted in the atmosphere. However, these are preliminary solutions and a more robust solution for sustainability at scale needs to be developed. Another important problem is standardization of communication protocols across the planet like cellular protocols where the constellations can effectively compete for the users without unnecessary augmenting overlapping infrastructure. On the UAV front, there are several traditional challenges of battery powered devices taken to the extreme – power, computation, storage, mobility and communication in both rural and urban environments.

As we move towards a more data-driven world, the role of AI becomes more inevitable. Especially with such mobile cyber-physical systems, it is critical to think about the guard rails and data-resilient technologies that remain safe for usage in human-coexistent environments. Further, parallel research in data sciences need to build tools and solutions that are trustworthy, reliable, and perhaps provable. An important aspect of that provable aspect will be the explainability of the model to describe what are the key features that had the most impact on the decision it took. As we learnt during a live talk by Olga Saukh, “the robustness of models to data drifts and sensors capability drifts will be critical to such sensors’ and data-driven algorithms’ ability to sustain equitable AI for cyber-physical systems.” It will also be critical to build solutions that understand and evaluate the data for data biases and build automated tools to understand the biases of black-box AI models. Further, human-in-the-loop models should create more accessible interfaces for the human to explain the deviation in the decision-making process that it is introducing and the logic behind it needs to be learnt by AI algorithms on-the-line. Many of the more energy hungry AI models will need energy-elastic capabilities to continuously manage the energy-accuracy tradeoff during live applications. Finally the continuous verification of these AI systems and detection of maleficent intent is critical for a safer future.

The impact of AI on IoT research has been significant and transformative. Integration of ML models into IoT systems has led to numerous advancements across various domains, including medical, underwater and space IoT applications discussed at the seminar.

As increasingly more data is generated by IoT sensors, ML algorithms running on edge devices extract valuable insights, identify patterns, and make predictions based on the collected data. This facilitates the development of new diagnostic tools and real-time decision-making. The approach is particularly valuable in manufacturing, but also builds the core of early warning systems in medicine, precision agriculture and cattle farming.

AI-powered algorithms can help to optimize network management and communication protocols in IoT systems. This improves connectivity, enhances reliability, and reduces latency issues. AI can play a crucial role in enhancing the privacy and security of IoT systems. ML algorithms can analyze network traffic patterns and detect anomalies that indicate potential cyber threats. They can help to anonymize and encrypt sensitive IoT data, ensuring privacy protection.

Natural language interfaces and vision-based tools can enable seamless interactions between humans and IoT devices, creating more intuitive and user-friendly interfaces and experiences.