Oswin So

Oswin So is a PhD candidate in aeronautics and astronautics whose research seeks to develop safe machine-learning methods for critical autonomous systems and bridge the gap between simulation and real-world systems. In previous research, Oswin has drawn on MathWorks’s Simulink to develop safe controllers for cases where the dynamics may be complicated but are known. As a MathWorks Fellow, he will build on this work, applying tools from reachability analysis and reinforcement learning to develop techniques that can provide safety guarantees even in cases where modeling errors are present. Concurrently, he will pursue new methods to perform adaptive safety by combining robust safety with online adaptation techniques such as adaptive control and Gaussian processes. Oswin has made significant contributions to the modeling and design of complex systems, drawing on MATLAB and Simulink, and enabled the development of advanced control algorithms that are both reliable and efficient. His research holds strong potential to deliver new tools and methods to advance real-world, safety-critical systems.