Allen M. Wang

Hi! I’m Allen, and I’m a fourth year PhD student at MIT working on controls-y and ML-y things for fusion experiments, specifically tokamaks. I’m co-advised by Cristina Rea and Chuchu Fan and am primarily based in the Plasma Science and Fusion Center (PSFC), although I am also affiliated with the Laboratory for Information and Decision Systems (LIDS). My home department is AeroAstro where I’m working towards a PhD in Autonomous Systems with a minor in Plasma Physics. I primarily see myself as a translator of fusion problems into computer science problems, so that the latest and greatest of the robotics and ML worlds can be applied to solve fusion problems.

My PhD is primarily focused on the problem of disruption avoidance: how should the tokamak make actuation decisions to avoid states that are vulnerable to plasma instabilities? It turns out simulating plasma dynamics is hard, and we don’t seem to have enough experimental data for standard seq2seq modelling techniques to solve all of our problems. Thus, a big part of my work is on developing hybrid physics-ML dynamics models, where I draw a lot upon ideas from the Scientific Machine Learning (SciML) community (although I do most of my work in JAX-land). In my work with researchers at EPFL, I’m working towards experimental demonstrations of a RL policy actively avoiding certain known instability regions. In my work with Commonwealth Fusion Systems (CFS) on SPARC, I’m working on: 1) building tools for hybrid-physics ML dynamics modelling, and 2) developing RL and robust trajectory optimization tools to assist in the offline planning of trajectories.

I spent my first two years of grad school (2018-2020) in the Computer Science and Artificial Intelligence Lab (CSAIL) under Ashkan Jasour and Brian C. Williams where I worked primarily on motion planning under uncertainty for autonomous vehicles. During that time, I also helped MIT Driverless get on its feet by leading the motion planning and controls effort. I then took leave from grad school and spent two years (2020-2022) as a roboticist at Boston Dynamics working on a little of everything on the Stretch robot.

I went to college at Iowa State University (2014-2018) where I was a bit academically vagrant, graduating with majors in Aerospace Engineering, Applied Math, and Economics. I found myself in robotics via research in mathematical optimization, some summers at JPL, and random hardware/controls projects. An internship working on the gravitational wave experiment LISA Pathfinder foreshadowed a future hanging around physicists.