Issue 125 | Apr-Jun 2021

Living On The Edge

Some of the most exciting developments in information technology are taking place in the smallest devices. NUS School of Computing explores this realm under the leadership of Professor Tulika Mitra.


The Cyber Connector

Professor Tulika Mitra, Provost’s Chair Professor of Computer Science, is Vice Provost (Academic Affairs) at NUS. Prof Mitra is a leading expert on embedded computing systems in the Internet of Things (IoT) and has authored more than 175 articles in peer-reviewed conferences and journals. She holds multiple US patents and enjoys collaborating with industry research partners (e.g., ARM, CSR, Facebook, Huawei, Xilinx) for real-world impact.
The Internet has an edge, and there are interesting things afoot there. To be clear, this is not like the bleeding edge of science, but a literal edge that marks the boundary between the physical and cyber worlds. For Professor Tulika Mitra, Provost’s Chair Professor of Computer Science, and Vice Provost (Academic Affairs) at NUS, the edge of the Internet is where the action is. Besides being a leading expert in the embedded computing systems that provide the foundation for the Internet of Things (IoT), Prof Mitra is a prominent figure in a number of different areas, but the IoT area requires a bit of a primer. Broadly speaking, IoT is a network of objects — sensors and devices — equipped with software to collect data and share the same with other objects and systems in the network. It is so pervasive today that it does not make the news in the way that artificial intelligence (AI) and machine learning do. Today, we perhaps feel this more than ever, in the era of the TraceTogether token (a small battery-powered Bluetooth device intended to support contact-tracing efforts during the COVID-19 pandemic in Singapore).

Working with small devices with limited power, including smartwatches but also other wearables such as fitness trackers, is one of Prof Mitra’s specialties, which is one of those interesting stories taking place on the edge of the Internet. “My focus is on the device, or rather the computing system in the device. At present these are somewhat limited, in terms of the computing capability, and require access to the Internet 24/7,” says Prof Mitra. The problem here, she continues, is this can be costly, especially when you think of cameras connected to the Internet, which are sending a constant stream of footage across the system. Of course, this involves a lot of bandwidth and is not a particularly efficient way of getting things done. 

This is where Prof Mitra’s work at the NUS School of Computing gets into interesting territory, because she is championing a better solution that can do most of the processing within the device itself. While that might seem at odds with the premise of IoT and the way it works with both Cloud computing and AI systems, Prof Mitra has a plan. Basically, it involves increasing the computing efficiency of devices, such that they do not need to rely on the cloud to do most of the analytics, while remaining low-powered. “You need to bring the power of a server to the tiny device using a small battery,” says Prof Mitra. “We want to make data analysis directly on the device rather than doing everything over the Internet. Believe it or not, it takes more energy to send data over the Internet for analysis than to do it on the device itself.” 
Prof Tulika Mitra (right) with PhD student Manupa Karunaratne.

Hardware Defined by Software

In her High-performance Embedded Architecture and Compilation (HiPEAC) 2021 keynote address, Prof Mitra noted that smartphones — a ubiquitous presence in most societies today — in particular play into the idea of bringing the power of a server to a small device. This is especially so, considering the well-publicised fact that today’s typical smartphone has more processing power than the Apollo missions to the moon had on-board the shuttles. 

With regard to the smartwatch, another device whose rate of adoption has soared in recent years, Prof Mitra asserts that the ideal scenario would be one where the watch could do much more work independently of the smartphone. For example, when it comes to analysing data from a fitness app, the watch would process all the data itself, rather than sending it to the phone to crunch all the numbers. Of course, in very small devices like this, the limitations are in both the chip and the power supply. 

We want to make data analysis on the device rather than doing everything over the Internet. Believe it or not, it takes more energy to send data over the Internet for analysis than to do it on the device itself.

No Limits 

living-3On succeeding in computer science as a woman
Although I work in one of the most male-dominated fields in academia, I have to say I never let it bother me. I got used to being the only woman in all meetings and committees. But we should do better for the next generation. There has been increasing awareness about diversity in my research communities in recent years and I am privileged to have the opportunity to actively participate in the initiatives supporting women colleagues.

On diversity and meritocracy
I don’t think there is a conflict between having diversity and being meritocratic. The quality of scholarship will always be judged by the same yardstick; that’s the cornerstone of academic life. Supporting diversity is about ensuring equity and inclusion by identifying and addressing any systemic barriers to individuals in the community in reaching their full potential. 

On leveling the playing field
In NUS and Singapore, we are very fortunate that women are not treated differently; the system is very fair. But, across the world, being small in number, women academics do not have access to the same social network as their male counterparts. These connections help you grow and make you aware of different opportunities. Apart from that, mothers with young children face unique challenges. Finally, women are not good at self-promotion — many of us experience impostor syndrome at different points in our lives. It is important not to forget about people who do not put themselves forward.

On Representation in Academia
The more successful women we have in academia, the easier it will be to attract women students and faculty members. The academia needs the role models. I am very fortunate because out of the five PhD candidates I recruited recently, three are women; I’m very excited to have this gender balance in my research group.
Prof Mitra thinks that one way forward for such devices is software-defined-hardware (SDH), or software that enables the reconfigurability of hardware. This forms part of her team’s work with the National Research Foundation (NRF) Competitive Research Programme on next-generation IoT edge computing. Edge computing here refers to “a distributed computing topology in which information processing is located close to the edge — where things and people produce or consume that information,” according to one definition of the term, by research firm Gartner. Basically, to continue with our analogy here, the smartwatch fits this definition to a tee. 

Obviously, despite the label ‘edge computing’, this is hardly a fringe area, with plenty of interest, both public and private. In the United States, the Department of Defense is studying SDH, in particular, under the Defense Advanced Research Projects Agency (DARPA), famous for being Internet and virtual reality pioneers. One of the chief benefits of SDH is its efficiency and adaptability. To elaborate a little, SDH will enable any given device to adapt to resolve new problems with existing algorithms, or use new algorithms to work on existing problems. Prof Mitra also notes that SDH has the potential to lower costs, while improving performance in terms of how many operations the device can perform per watt of power. 

Seeking Solutions that Matter

What is particularly interesting with the Competitive Research Programme is that it involves partners who have real-world problems to solve. One example of this is a project Prof Mitra is collaborating on with A*STAR’s Institute of Infocomm Research (I2R) alongside the Institute of Microelectronics (IME). A compelling real-world scenario is to enable security officers to make dynamic sweeps of events, when looking for suspicious activity or persons. The original toolkit involved a cumbersome device in a backpack that only had a battery life of one hour, among other things. Prof Mitra and her team are working on a solution in a smartglass form-factor that will last for about three hours using tiny batteries. “To succeed in terms of the performance per watt here, we will need to be 20 times over where the technology is right now. So this is an end-to-end system that we are very excited about,” says Prof Mitra.

Succeeding in addressing real-world problems is something Prof Mitra has experience with, given that her research has been supported by both public and industry funding. “I’ve been fortunate to have public funding that has been complemented by industry funding. While the money is important, the biggest benefit for me is that industry partners can tell you upfront that the assumptions you have are wrong! I mean, I can have a beautiful solution in mind but would anyone care? It can be great curiosity-driven academic research, but it might not solve any real problems,” says Prof Mitra.

Having said all that, solving real problems is something Prof Mitra has to do on a daily basis because she is incredibly busy. Aside from teaching, research and administrative duties, Prof Mitra also organises numerous conferences, including most recently when she served as the General Chair of ESWEEK (Embedded Systems Week) 2020 and the Program Chair of ICCAD (International Conference on Computer-Aided Design) 2021, the premier events in embedded systems and design automation, respectively. She also holds various editorial positions in leading computer science journals, making time a precious commodity for her. Prof Mitra confesses that a typical day for her “looks bad and is very long”. 

Text by Ashok Soman.
Load more comments
Thank you for the comment! Your comment must be approved first