Jonas Hall: Implementing an Optimization Solver Anyone Can Use
By Chloe Cramutola
He started as a mathematician, but his engineering interests took him beyond basic numbers.
“For the longest time, I was [mostly] interested in pure mathematics,” said Jonas Hall, a third year PhD student in Systems Engineering at Boston University. “I didn’t really [appreciate] applied mathematics, but I couldn’t find a field where I felt at home.”
He received his bachelor’s in Mathematics in 2018 from the University of Freiburg. Although he took one year to study in the Baden-Wuerttemberg graduate exchange program at the University of Massachusetts, he came back to Freiburg to get his master’s degree.
“During those two years I worked closely with [Professor Dr. Moritz Diehl] in the optimization and controls community,” said Hall. “He was [one of] the most inspiring professors I’ve ever met.”
Now at Boston University, optimal control is Hall’s main focus. In optimization, the aim is to find a minimum of a function over a feasible set. In control, he works on manipulating a system so that he can achieve stability or achieve certain tasks.
When control and optimization are combined, the desired goal is to control the system in an optimal way. This is an engineering design technique used to control and optimize complex systems.
One tangible example is the moon lander, which proposes the question: How can this spacecraft land on a lunar surface, all while using the least amount of fuel? Hence, the goal is to minimize the fuel and time it takes to land simultaneously.
For Hall’s PhD project specifically, his work includes the persistent monitoring problem.
This problem arises when cooperating mobile agents (agents that keep data intact as they migrate from one machine to the next) must monitor a changing environment. These agents can pick up where they left off and perform necessary tasks in the new environment. Tasks can include coverage control, surveillance, and environmental sampling.
“We’re interested in trying to cover or monitor points of interest or a certain area with autonomous agents––so, robots,” said Hall. “You essentially assume the environment is dynamically changing so that you have to move around the environment, in order to monitor everything persistently.”
However, Hall started a six-month summer internship this past June at Mitsubishi Electric Research Laboratories (MERL). This is quite different from his daily research, as it focuses less on control and more on optimization.
“My PhD environment is very self-paced, and I had a lot of time to think about the problems that I’m working on,” said Hall. “Here at the internship, it’s much higher-paced and I was more out of my comfort zone. It was difficult at times, but I was excelling at a different speed.”
Hall is working on implementing an optimization solver specifically for quadratic programs, which is a subfield of convex (curved outward like the outside of a circle) optimization problems. These problems can be feasible or infeasible.
Mitsubishi Electric tasked Hall with developing an open source solver that people can use for various problems (like the persistent monitoring problem). Essentially, it’s just code that anybody can contribute to and use.
“So that’s something that we will release and then we will see how the community will adapt it whether there is use for it or not,” said Hall. “Hopefully there is and the way that people would use it is that they would formulate an optimization problem.”
Ideally, people would use this specific algorithm so they can solve their formulated problem and certify infeasibility or optimality.
Hall explained that it sounds abstract, but problems arise in many types of applications, such as in power grid systems or reference tracking (as used in autonomous vehicles).
“[The internship] was also a great opportunity to meet many new collaborators and friends,” said Hall.
Thanks to a friend from the University of Padova in Italy, he was able to set up a new internship more related to his PhD. He set up a collaboration between his current BU advisor, Mechanical Engineering Department Chair and SE Professor Sean Andersson, and Associate Professor Ruggero Carli at the Department of Information Engineering.
While Hall is planning to graduate with his PhD at the end of his fourth year, he’d also like to give back and educate students like himself.
“I’m really interested in teaching. I really enjoy building connections with humans and trying to pass on knowledge,” Hall said. “I think that’s a very peaceful thought. I do also enjoy the research aspect. But I think ideally, I’m looking for something that is more teaching focused with a small load of research on the side.”