Matthew C. Fontaine

Thanks for stopping by!

I’m a researcher dedicated to algorithmically generating scenarios to evaluate and test robots and AI. A major goal of my work is to help us understand these systems, find dangerous faults, and develop rigorous testing methods for modern AI.

A major thrust in my research has been quality diversity optimization, a class of optimization that finds a diverse collection of solutions, while also maximizing an objective. During my PhD, we developed the Covariance Matrix Adaptation MAP-Elites (CMA-ME) and Covariance Matrix Adaptation MAP-Annealing (CMA-MAE) algorithms. We introduced Differentiable Quality Diversity, the first-order counterpart to blackbox quality diversity optimization. All of the algorithms are implemented in the pyribs library.

My research leverages quality diversity optimization as a search method to find good coverage of interesting failure modes in autonomous systems. Together with my labmates in the ICAROS lab at the University of Southern California (USC), we developed a general framework for finding failures against arbitrary blackbox robotic systems.

I’m now looking for avenues for continuing my research. If you have an opportunity for me, please don’t hesitate to reach out!