The Robotics Institute
Human and Robot Partners Lab
I am a Masters (by Research) student and a Graduate Research Assistant at the Robotics Institute at Carnegie Mellon University. I work in the Human and Robot Partners (HARP) Lab and the Transportation, Bots, and Disabilities (TBD) Lab. I am advised by Prof. Henny Admoni and Prof. Aaron Steinfeld, on robot self-assessment, specifically robot perception and reasoning for predicting the effects of robot interaction with the physical world.
- Proposed object-conditioned multi-view CNNs to predict, a priori, the resulting stability of partially occluded cluttered scenes and guide a robot in safely extracting objects.
- Developed a data-simulation pipeline using MuJoCo to generate synthetic data for training the model and used domain randomization to transfer it to real-world scenes.
- Delivered a real-time pipeline for object detection using Mask-RCNN and accomplished object-correspondence across multiple-views using deep feature matching and completed its visual assessment using the proposed model.
- Successfully demonstrated an overall reduction in the manipulation failure rate around careful object selection by 15% on our collected dataset (under review at IROS 2020). Read the abstract for this work here. Manuscript is available on request.
- Member of the SUCCESS MURI. My work is supported by the Office of Naval Research (N00014-18-1-2503).