On October 7, Gregory Chirikjian, Willis F. Harrington Professor and Chair of Mechanical Engineering at the University of Delaware held a seminar titled “From State Estimation on Lie Groups to Robot Imagination” at Yerevan State University. 

The seminar spanned multiple topics involving probability, harmonic analysis, and geometry as applied to robotics. Professor Chirikjian discussed how modern robots can overcome their limited intelligence by learning to "imagine": scanning unfamiliar objects, simulating their potential uses, and then applying the learned affordances in real-world actions. In this way, robots can reason about objects that they have not encountered before. After affordances are assessed, the robot then takes action in the real world, resulting in real2sim2real transfer. As part of this broad framework, probabilistic methods on Lie-groups are used. These mathematical methods were developed originally by the presenter for mobile robot state estimation, and have been adapted recently to one-shot learning of affordances from demonstration. 

The seminar also featured engaging demonstrations and future plans for integrating large language models into robotics.  

The participants actively engaged in discussions and asked many questions as this was a unique opportunity for them to speak with a professional and gain a broad overview of robotics.