My project experience is across the fields of computer vision, designing aerial and indoor robotics software systems, planning for robotics, embedded systems, electronic design automation (EDA), mobile app development and web development. It has been acquired through working in industry, academic research, startups and through participating in numerous competitions.

June 2019 - June 2020


Pose Estimation of Moving Objects on a Conveyor Belt

Location: Search Based Planning Lab, RI, CMU

PR2 is a full body mobile manipulator (or humanoid) that consists of a 7-DoF arm. In this project, I implemented an ICP based pose estimation strategy that works with a search-based planner to grasp moving objects of a conveyor belt. I improved the pose estimation strategy by deploying PERCH 2.0 for this task enabling spatially consistent accuracy across the conveyor length.

Papers : RSS 2020, Master’s Thesis / Videos : Clip 1, Clip 2 / Code : GitHub / Skills : Python, ROS, PCL

Oct 2019 - Dec 2019


Improved MR-MHA* using Conditional Variational Auto Encoders

Location: RI, CMU

Developed a combined learning and search based planning method that reduces planning time by using prior experiences learnt through CVAEs. The planner used (MR-MHA*) exploits loose coupling between high dimensional state representations such as between the base and arm of a mobile manipulator and CVAEs are used to better exploit this loose coupling based on prior experiences.

Links : Report, Presentation / Videos : Clip 1 / Skills : Python, C++, ROS, Planning

Feb 2019 - Oct 2019


Perception and Grasping for Crate Opening on RoMaN

Location: Search Based Planning Lab, RI, CMU

RoMaN is a mobile manipulation platform developed as part of the RCTA project which is a collaboration between several universities, ARL, JPL and GD. In this work, I deployed perception via search for crate detection on RoMaN. We demonstrated higher accuracy as compared to pure learning based methods and demoed the full pipeline of pose estimation followed by crate opening at NREC. The demo was attended by eminent defense personnel.

Papers : SPIE / Videos : Clip 1, Clip 2 / Media : / Skills : C++, ROS, PCL, OpenCV, MPI

Sept 2019 - Present


Full Body Mobile Manipulation on Walker

Location: Search Based Planning Lab, RI, CMU

Walker is a full body mobile manipulator (or humanoid) developed by UBTech that consists of a 3-DoF mobile base and a 7-DoF arm. In this project, I co-wrote the software stack for the entire robot, comprising of planners (arm and base), state machines, controllers and perception modules enabling it to navigate indoors and fetch objects for the user. The capabilities of the robot were demonstrated to over 50 people as part of a continuous demo day in July 2019.

Videos : Clip 1 / Code : GitHub / Skills : Python, ROS, PCL