Paloma Sodhi
psodhi at cs dot cmu dot edu

I am a final year PhD student at The Robotics Institute at Carnegie Mellon University advised by Michael Kaess. I’m interested in learning and optimization methods for robot perception — How should robots efficiently infer latent physical states from a stream of complex measurements to make better decisions?
In my Ph.D., I developed algorithms for learning deep observation models and optimizing with physics constraints in factor graphs. I have applied my work to real-world tactile manipulation and underwater navigation applications. I actively collaborate with Facebook AI Research and am a recipient of the CMU Presidential Fellowship.
Previously:
I completed my Master’s in Robotics at CMU working on 3D computer vision methods for plant phenotyping. I hold a Bachelor’s in Electrical Engineering from IIT Guwahati, India.
Recent News
Mar '21 | Our work on learning tactile models for factor graph-based estimation accepted at ICRA'21. Preprint available on arXiv. |
Dec '20 | Successfully completed my thesis proposal! Talk available here. |
Dec '20 | Excited to present our work on learning tactile observation models in factor graphs as a contributed talk at the WiML workshop, NeurIPS'20. |
Aug '20 | I spent a lovely summer at Facebook AI research working with Stuart Anderson and Mustafa Mukadam on tracking object states using tactile image sensors during robot manipulation. |
Jun '20 | Excited to present our work on incremental constrained smoothing (ICS) in factor graphs at ICRA'20. Check out our paper and talk! |
Nov '19 | Check out our work on online and consistent occupancy mapping with factor graphs at IROS'19 paper, IROS'18 paper. |