Aditya Kamireddypalli

I am a final-year PhD student in Robot Learning at the University of Edinburgh, advised by Subramanian Ramamoorthy and Steve Tonneau. My research focuses on enabling robots to reliably perform contact-rich manipulation tasks under uncertainty.

I am particularly interested in how robots can use physical interaction as a source of information. While vision provides broad information about the environment, contact provides local and precise information about geometry, pose, and the state of an interaction. My research explores how these complementary sources of information can be combined to improve both perception and control.

A central part of my work is probabilistic inference for robotics. I use Monte Carlo inference methods, including Markov chain Monte Carlo (MCMC) and Sequential Monte Carlo (SMC), to represent and reason about uncertainty. I am particularly interested in combining probabilistic programming languages with simulation-in-the-loop, allowing physics simulators to serve as generative models for inference from visual and contact observations.

I also study how the resulting uncertainty estimates can directly inform robot actions. My work combines probabilistic inference with sampling-based planning and control, using contact information not only to estimate the world but also to adapt how a robot explores and interacts with it.

Overall, my research aims to close the loop between perception, physical interaction, probabilistic inference, and control to make robotic manipulation more robust under uncertainty.