Mitchell B
- Research Program Mentor
PhD at University of Michigan - Ann Arbor
Expertise
Robotics, control systems, AI/machine learning
Bio
Focusing mainly on control and estimation for robotics, my research tends to explore methods that provide safety and performance guarantees in pursuit of forging new frontiers in autonomy. My aim is to develop novel control, estimation, and machine learning-based techniques toward advancing our collective understanding of the robotic mind and its ability to interact with bodies both human and otherwise. I feel that there is tremendous potential still to be unlocked in the field of robotics/AI, and every day I strive to contribute to this endeavor. Consistent with my interest in robotics of sound mind and body is my love of running, indeed there are few hobbies that demand so much of the two. Outside of the lab, you may find me striding up one of Ann Arbor’s many hills or clipping off quarter mile repeats at Michigan’s storied Ferry Field. An accomplished middle-distance runner during my time at Tufts University, I competed at the 2021 US Olympic Trials 800m in Eugene, OR.Project ideas
Design of a Human-Powered Flying Machine
A student would investigate some of the physics and the mechanisms that allow birds to fly. A comparative analysis would then be conducted between various birds to discover why some are better flyers than others (e.g. peregrine falcon vs. turkey). Using this knowledge, the student would then design a theoretical device to enable human-powered flight.
Certifiable Safety for AI / Machine Learning Systems
Artificial intelligence (AI), and in particular new advances in generative AI, are quickly becoming ubiquitous in our society. From search engine augmentation to robotics, AI models like ChatGPT are being used to solve challenging problems that not long ago were seemingly impossible. But how do we know that we can trust the outputs of these AI systems? In this project, you might investigate how to evaluate the correctness of fundamental models, and develop mechanisms for rectifying erroneous or even unsafe model outputs in order to promote further trust and confidence in AI. Example applications include image classification, object detection, climate modeling, robotic system control, etc., the possibilities are endless!