Year in Review: 2020

Summary: A copy of my end of year message to my lab, where I review the key events from 2020.

As 2020 draws to a close, I think we can all say that this year took on quite a unique flavor compared to last year in every sort of way. That said, there are many things that remained the same, including the drive and commitment that all of you have shown under such exceptional circumstances, given both my own paternity leave at the beginning of the year and then later when dealing with the pandemic and its remote work consequences. I am grateful to have had such a wonderful and gifted group of collaborators to help weather the collective storm that was 2020. Despite these difficulties, we have had some genuinely bright spots throughout this year to celebrate, which I will try to highlight below for you, as well as some key things to look forward to in 2021.

The People

As I have mentioned before, the best part of the lab for me is all the wonderful people I get to collaborate with and call our collective lab family. This year marked an important year of renewal for the lab, as we had several departures and additions that formed part of the natural ebb and flow of academic research labs. I must admit that I am still getting used to that process, as I find it difficult to see good friends and colleagues move onto new things, but it is also exciting to bring in new people with fresh approaches and ideas. I missed doing many of our typical retreats and lab activities this year, on account of pandemic restrictions, but am looking forward to starting those up again safely in 2021.

First, over the summer Rachel Hess graduated with her M.S. thesis entitled “Automatic Optimization Methods for Patient-Specific Tissue-Engineered Vascular Grafts.” Rachel worked with Axel Krieger and I (along with our collaborators at UChicago and Children’s National Hospital) for our Design + ML related NIH project on designing 3D electrospun grafts for children born with congenital heart disease. She has since joined the FAA nearby working in their trajectory optimization group. Second, both Dr. Charlie Manion and Dr. David Anderson finished their contracts with UMD over the summer and are on to new endeavors (including a large round of angel investor funding for David’s design automation startup). I will miss all of our great conversations and explorations that we had as part of the DARPA Fundamental Design program, which has given me at least 5-10 years of new proposal and paper ideas from all the places we got to explore together. Third, Dr. Xiaolong Liu has transitioned to a Research Scientist role at Johns Hopkins, though we are still collaborating on our active NIH grants and Xiaolong was pivotal in securing a follow-on grant coming in 2021 that I cannot publically announce yet, but I am looking forward to working on it with him.

In terms of additions, this year we were joined by three excellent Ph.D. students, Eesh Kamrah, Qiuyi Chen, and Xuliang Dong, who will be supporting several NSF and ARPA-E projects, as well as an M.S. student, Richard Moglen, who will be looking at joint work in Soft Robotics with post-doc Fatemeh Ghoreishi, and PIs Ryan Sochol and Axel Krieger. I have already benefited from many of the great skills, perspectives, and energy that they have brought with them into the lab; especially given that their “first-year graduate experience” during a pandemic is unlike anything others have had to go through. Truly impressed!

The Science

This year we made a bunch of progress across many scientific areas, with almost too many applications to exhaustively mention here. Some of these were closing the loop on existing projects from 2019, though many also represented new endeavors that we’ll see bear more fruit in 2021 and beyond. I’ll highlight a few common threads below that provide exemplars of this trajectory, though I think you can get a broader sense of our entire research portfolio by checking out our papers page or Google Scholar. For example:

Wei spearheaded the publication of a culmination of his work on Design Manifolds via an AIAA Journal paper finally published this year called “Airfoil Design Parameterization and Optimization Using Bézier Generative Adversarial Networks” as well as the ArXiV preprint for “Adaptive Expansion Bayesian Optimization for Unbounded Global Optimization.” These two papers brought together years of work that he has done in his thesis on the usefulness of low dimensional “Design Manifolds” for exploration and optimization. In his most recent work, we show how design manifolds can accelerate airfoil optimization methods 2-3x that of State of the Art, and up to 10x that of popular methods. It laid the foundation for work in two of our new ARPA-E projects, so I hope to share more about how we built upon it in the coming year. As of writing, the code for that paper and the underlying BezierGAN architecture code are currently our most popular repositories on our lab’s GitHub. So I’m really excited to see where we and others can take some of these ideas, as it represents the real culmination of the work we did under the DARPA Young Faculty Award program mentored by Jan Vandebrande.
Faez finalized several publications that came out of our joint work with collaborators Scarlett Miller and Sam Hunter at Penn State on learning and optimizing metrics for design variety and creativity measurement, such as “Design Variety Measurement Using Sharma–Mittal Entropy” and “How should we measure creativity in engineering design? A Comparison of social science and engineering approaches”. The types of problems we explored in those papers, as well as his earlier work, had been gnawing at my brain ever since I was a graduate student and so it felt nice to finally get our hands around generalizing some of those concepts in a more unified way. This led to ongoing work that I hope to see come to fruition in 2021 or 2022, so stay tuned.
Jun and Nicholas continued putting out work that came out of the DARPA Fundamental Design program, such as on the Learning to Abstract and Compose problem. We explored so many interesting directions in that program and have more papers in the pipeline to release in 2021.
Dan’s 2019 review paper on “Deep learning for molecular design” in Molecular Systems Design & Engineering by the Royal Society of Chemistry continues to gain traction in the growing ML + Materials Design community, so I am glad that it is remaining relevant in such as rapidly evolving area.
Faez’s continued collaborations with Saba Ahmadi, John Dickerson, Samir Khuller and I on various applications of optimal diverse matching has produced multiple new avenues such as diverse team selection and multi-attribute matching. This formed some of the foundations for one of our new ARPA-E collaborations that John and I now have underway.
Mark, along with Co-PI’s Ken Kiger, Miao Yu, David Bigio, and Steve Michell received a Provost Teaching Innovation Grant over the summer to help re-tool our department and college’s teaching tools by integrating and deploying a system called PrarieLearn (developed out of UIUC) for mastery-based teaching. It turns out that this grant was actually the largest one that the MechE department received out of the Provost’s initiative, to my surprise. A number of you in the lab helped get this off the ground over the summer, including David Anderson, Charlie Manion, and Nicholas Chiu, so thank you all again for your help on this. It ended up being widely used for over 200+ students in the Fall semester with more coming in the spring.

While this was just a brief snapshot of some of the outcomes this year – and is by no means exhaustive – I wanted to use it as an opportunity to show you all how the work you are doing today really does lay the groundwork for yourselves and future students to build upon. Much of what we do and celebrate today would not be possible without seeds that students planted many years ago and dutifully tended until today; so you can see your own trajectory as forming the life-blood of the lab itself.

I also wanted to remind everyone that all of the above would not have been possible were it not for the support of our various research sponsors that, at various parts of 2020, included seven projects across DARPA (via the FUN DESIGN and YFA programs), NSF (via the EDSE program), the NIH (via the NHLBI), and ARPA-E (via the DIFFERENTIATE program). So I do thank you all for your many efforts – big and small – to help deliver on the promises we make to these sponsors and the missions they support.

The Field

There continues to be broad interest in the intersection of Machine Learning and Mechanical Design/Optimization. For example, while 2019 saw a Journal of Mechanical Design Special Issue on “Machine Learning in Engineering Design”, JMD also recently announced another SI in “Artificial Intelligence and Engineering Design” which includes ML as a subset. I have also noticed beyond our own lab both a growth in the number of ML + Design related publications in our major conferences, and, perhaps more importantly, more nuanced application, critique, and contributions of those articles beyond what I had seen even five years ago. I think this speaks to a growing set of researchers who’ll be putting forward great work in these areas in years to come, and I am excited by this growth and interest. As always, if you hear of any up-and-coming researchers that you think I should add to my reading lists, let me know since I’d like to support and keep abreast of the best work in this growing area.

On the downside, the COVID-19 pandemic pushed back our plans for an in-person 2020 NSF Workshop/Summer School on Emerging Mathematical Foundations for Engineering Design, however, we’re gearing up to offer this in 2021 and investigating methods to do it in a safe and engaging way, so I look forward to sharing updates with folks on that as the world and our plans evolve out of this pandemic.

Funding wise, this year saw the launch of ARPA-E’s DIFFERENTIATE Program, which is targeted at using ML to improve the design of energy products. Our lab is involved in 3 of the 23 funded efforts, which spanned a dizzying array of application areas from what I could tell at the kickoff meeting. I was really impressed by the range of ML + Design applications that various industrial, academic, and national lab partners were involved in, and this gives me great hope that there is a real commercial and societal impact to be made by research investments in this area. I think what this means to you as lab members are (1) that, if done rigorously and on important problems, your research can have a substantial real-world impact; and (2) that I expect the employment opportunities for folks with our types of skills at this intersection will remain strong at least over the next decade or so. I’ve always enjoyed the academic research angle of ML+Design type questions (thus why I still run an academic lab), but it has become abundantly clear to me now that industrial partners, national labs, and startups are also active and growing in this space. What a great time to be working in this type of research! Look out next year since we have some new grants/projects coming in from sponsors like ARL, ONR, and two others that are yet to be publicly announced, so I am excited about what those possibilities will enable for us as well.

Alumni News

Occasionally, I’ll get some updates from past lab alumni that I like to share with the lab family each year. Specifically, Faez officially kicked off his lab at MIT this fall, so we all wish him well at a strong start there. Wei has been leveraging some of his research skills at deploying new internal projects and tools at Siemens, so it is nice to hear that our work remains industrially relevant. Zois is settling into American University and is helping spearhead the initiation of a new Ph.D. program there, so that has been an exciting opportunity for him to shape the future of his department and university. Thanks to all the alumni who occasionally report back on their changes over the year so that I can share many of your updates with those who follow in your footsteps.

Thanks everyone again for what you have been able to accomplish in spite of a difficult 2020, and I look forward to what 2021 has in store for us!