Invited Speakers

Mobile IoT plays a critical role in collecting heterogeneous data in a geo-referenced and granular manner, supporting data collection for various scientific disciplines. SenseMyCity is an iteratively designed mobile opportunistic IoT crowdsensing tool that evolved organically to meet the growing requirements of interdisciplinary sensing projects, from biomedics and psychology to transportation and networking. Over the years, we collected around 100 Million GPS data points from more than 1000 participants, spanning more than 550000 km and 60000 hours of GPS trajectories.

This talk will show 3 flagship projects enabled by SenseMyCity, as well as capabilities that were iteratively added to the tool to improve usability. First, it was used to collect 145h of daily work-trips from 36 bus drivers, including their cardiac signals, facilitating memory recall of stressful situations and contributing to characterize bus drivers' cardiac stress sources. Secondly, SenseMyCity's capability to automatically collect high-frequency sensor data only during users’ trips allowed us to quantify the advantage of using public transportation, and we were able to monitor the impact of a specific policy change. Finally, we propose to use granular human mobility profiles for informing offloading strategies during commuter trips.

Learning the evolution of public opinions around a topic, a.k.a. opinion dynamics, has multiple applications. In today's time, often, opinions expressed on a public or broadcast forum are formed through interactions over multiple online and offline platforms. We consider the problem of learning the influence graph and the dynamics using only binary opinions expressed on a public or broadcast forum, for example likes/dislikes, sharing, re-posting. We study a model where the private beliefs evolve according to the well known linear dynamics on an influence graph and the opinions are the binary expressions of those private beliefs. However, the learner has access only to the binary opinions expressed at different times. We propose a polynomial time algorithm which is provably accurate and provide a guarantee on the sample complexity under some mild conditions.

In this talk, we discuss mathematical modelling of the BBR congestion control algorithm, recently developed by Google. BBR differs from loss-based TCP—which uses packet loss as a measure of congestion—in that it uses an estimate of bottleneck bandwidth and round-trip time to infer network congestion. Owing to this fundamental difference, existing fluid models for TCP cannot be used for analysing networks that operate over BBR. This necessitates the development of a more evolved fluid model for the underlying end-to-end feedback system. In our work, we develop and propose such a fluid model for BBR operating with routers designed as per (a) large and (b) small buffer-sizing regimes. Further, we validate the proposed model through simulations conducted on NS2. The potential use of the proposed model for a system-theoretic analysis will be outlined, and its appropriateness will be highlighted.

The introduction of IoT in healthcare has led to many advancements and has added new dimensions to the traditional healthcare systems. The interactions between patients and doctors have moved beyond physical visits with the integration of IoT in the healthcare domain. Digital storage of medical data, remote health monitoring, and historical data analysis are some of the significant developments that have been accomplished in the area of healthcare. Wearable devices have played a vital role and have proved to be a game-changer in embedding IoT in healthcare. In the present scenario, wearables have witnessed growing popularity and have become a part of people's lifestyles. Wearables have enabled continuous monitoring of physiological parameters, both locally and remotely. Monitoring the elderly, children, and patients with chronic illnesses have especially benefitted with the availability of wearables. In addition to this, more complex systems for the diagnosis of diseases provide better insights into the patient's condition, which helps in early detection, faster response time, and targeted treatment. IoT-enabled smart healthcare connects patients from less accessible rural areas to remote healthcare professionals. In this lecture, we will discuss the state-of-the-art in the healthcare domain and its evolution with the introduction of wearables and diagnostics systems.

In this age of vaccines and antibiotics there is still a constant effort to find new drugs to combat illness for which there are no known cures. There is a need to discover replacements for existing drugs for pathogens that have become resistant to current drugs. In addition there is a need to develop new drug therapies for health issues that adversely affect the lives of hundreds of million people every day of every social standing in every population. Antibiotics and similar drugs have been used so widely and for so long that the infectious organisms the antibiotics are designed to kill have adapted to them, making the drugs less effective. Although the multidrug-resistance in pathogens is growing fast, the number of new drugs being developed to treat bacterial infections has reached its lowest point since the beginning of the antibiotic era. Hence, there is a dire need to develop new platforms and approaches to discover antimicrobial agents against novel molecular targets. Not only new drugs are not being created, but the existing process of creating drugs is slow, inefficient and costly. To address the issues of multidrug-resistance and newly emerging pathogens, we have developed Deep Drug, a computer aided drug design software utilizing AI-based techniques that access very large datasets and, in turn, creates an improved method for identify new compounds rapidly thereby dramatically shortening the developmental time line and the cost to have a new drug accepted.

Extended Berkeley Packet Filter (eBPF) is a mechanism that emerged recently in Linux to extend the functionality of the operating system kernel. eBPF works by downloading user code into the kernel and executing it at specific points of attachment within the kernel---for example, the network device driver. By skipping irrelevant kernel code paths and running the code directly in the kernel with minimal run-time checking, eBPF provides significant performance advantages over a user-space implementation. To mitigate security risks from running untrusted user code, the kernel implements software fault isolation through static program analysis, encapsulated into an in-kernel component called the verifier. If the verifier cannot prove that a program is safe, the program is rejected and cannot execute. Hence, eBPF combines the trifecta of flexibility, safety, and high performance in a familiar Linux environment. eBPF code is already widely deployed in several production systems.

Despite the promise, programming high-performance networking applications in eBPF creates several new challenges. Optimizing compilers need to incorporate safety into their optimizations. Developers must wrestle with the arcane rules of the verifier to prove the safety of their code. The verifier's static analysis algorithms contained critical bugs, resulting in disastrous security consequences. This talk will cover our recent work to address some of these challenges.

Opinion dynamics is a topic of interest in various areas of research including mathematics, physics, social sciences, economics etc. In the era of social media, our views are increasingly influenced by the online groups or communities that we are part of. We are also constantly exposed to various forms of advertising. In this context, there is a growing interest in opinion formation and influencing strategies. In this talk, I will introduce an opinion dynamics model on finite graphs evolving over a finite time horizon. The population has binary opinion and the individuals exhibit conformist as well as rebel behaviour. Assuming that the external influencing agency has a fixed budget in terms of how many time-slots can be influenced, I will discuss optimal opinion-shaping strategies depending on the behaviour of individuals (which may change with time) and the underlying graph structure.

Joint work with Sharayu Moharir, Bhumesh Kumar, Arunabh Saxena and Anmol Gupta.

Emerging applications such as smart cities, autonomous vehicles, and mixed reality systems rely on embedded systems that are engaging with the physical environment through sensors. Building upon this connection, my vision is to advance Omnipresent Sensing by harnessing the wireless infrastructure in and around buildings and cities to act as a non-intrusive sensing platform. This is possible by innovating at the crossroads of two recent trends in mobile computing: (1) Wireless technologies such as Millimeter-wave (mmWave) and Massive MIMO systems can now support higher bandwidth for communication and improved resolution for RF sensing applications. (2) Advancements in CPUs and RF front-ends are making it easier to develop software-defined sensing and communication systems. This is affording edge devices the ability to do more advanced signal processing and machine learning. In this talk, I will focus on how to design an RF-equivalent of optical retro-reflectors and use them as passive fiducial markers in autonomous vehicles, robotics, and mixed reality applications. I will then discuss how nuances from the environment itself can be leveraged to improve sensing quality in the context of human sensing, object tracking, and indoor localization. I will conclude this talk with a roadmap of combining radar-style RF sensors and wireless communication links for the wireless embedded systems of the future.

Emerging technology including AI and IoT builds strong momentum for studying novel applications. Research on applications touches on many exciting challenges. Besides technical improvements, we need to understand what practices and problems people currently have, and how we can design technology to better fit their activities. Application-oriented thinking enables researchers to develop a capability to obtain a comprehensive perspective on both technology and humans in addition to skills to drive technical breakthroughs. In this talk, I present several recent projects in sensing technology for interactive systems and applications, and share the joy of application-oriented research.

The ability to communicate and sense across the air-water boundary is essential for efficient exploration and monitoring of the underwater world. Existing wireless solutions for communication and sensing typically focus on a single physical medium and fall short in achieving high-bandwidth communication and accurate sensing across the air-water interface without any relays on the water surface.

In this talk, I will describe our effort in exploiting the use of laser light to enable effective communication and sensing in the air-water context. I will present our AmphiLight framework that allows bidirectional, Mbps-level communication across the air-water interface. I will focus on our design elements that overcome full-hemisphere laser steering with portable hardware, handle strong ambient light outdoors, and adapt to dynamics of water waves. I will also briefly discuss our ongoing effort on air-water sensing, and conclude with future work.