Carnegie Bosch Institute awards 2021 projects

The Carnegie Bosch Institute recently awarded its third round of interdisciplinary projects, which were chosen for research at the intersection of modern data-driven AI and classical scientific or engineering approaches. Whereas current AI methods leverage an abundance of information to learn data-driven solutions, classical methods rely on mathematical modeling of expert knowledge derived from the laws of natural science / engineering. Hybrid models seek to bridge these two approaches (AI + classical) by incorporating both data-driven methods and knowledge of the physical world, e.g. augmenting “traditional” simulation tools with AI-driven methods.

Funded projects include:

• Hybrid 2D-to-3D localization in changing environments, Sebastian Scherer (PI, Robotics Institute), Burcu Akinci (Co-PI, Associate Dean for Research, College of Engineering)
• Using out-of-sample regularization of physics-informed neural networks to speed up computational fluid dynamics, Alex Davis (PI, College of Engineering), Aarti Singh (Co-PI, School of Computer Science), Satbir Singh (Co-PI, College of Engineering)
• Scheduling and queueing algorithms for resource-sharing in federated learning, Gauri Joshi (PI, College of Engineering), Weina Wang (Co-PI, School of Computer Science)
• Safe reinforcement learning integrating physic laws, control theories, and formal, Ding Zhao (PI, College of Engineering), Conrad Tucker (Co-PI, College of Engineering), Eunsuk Kang (Co-PI, CyLab)