Speaker

Eric Ewing

Abstract

Recent advances in multi-robot systems have created opportunities for widespread deployment in applications ranging from warehouse automation to search and rescue. However, coordinating large groups of robots remains challenging, both in terms of computational efficiency and accessibility to non-expert users. This talk explores novel approaches to multi-robot coordination through the lens of Multi-Agent Pathfinding (MAPF). We first examine data-driven techniques for improving solution efficiency, including algorithm selection, hardness estimation, and instance decomposition methods that can improve the performance of a portfolio of algorithms. We then discuss an unexpected direction: using robot art as a means of increasing accessibility and public engagement with robotics, while simultaneously revealing new research directions through open-ended exploration. The talk concludes by examining future opportunities at the intersection of deep learning and optimization, investigating how we can leverage modern machine learning approaches to enhance planning while maintaining theoretical guarantees. Throughout, we focus on bridging the gap between theoretical advances in multi-robot coordination and practical, accessible deployment.

Bio

Dr. Eric Ewing received his PhD from Brown University in July of 2024 for his dissertation entitled "Advancements in Portfolio Methods for Optimal Multi-Agent Pathfinding" under the mentorship of Professor Nora Ayanian. Prior to Brown, Dr. Ewing studied at the University of Southern California (USC) before moving with his advisor. He is currently a visiting professor at Brown University, where he teaches Introduction to AI, Deep Learning, and a unique class "Robots as a Medium," in which students work on semester-long art projects using robots. His research lies at the intersection of modern AI techniques, like deep learning, and classical AI techniques of search and planning for large-scale multi-robot coordination problems.