Noushad Sojib

I am a PhD student at the Cognitive Assitive Robotics Lab Computer Science, University of New Hampshire, where I work on Imitation Learning. My PhD advisor is Momotaz Begum.

I have complted undergrade in Computer Science and Engineering from Shahjalal University of Science and Technology where I had the pleasure of working with M. Zafar Iqbal and Md Ruhul Amin.

Email  /  GitHub  /  Google Scholar  /  LinkedIn

Research

I am interested in making robot learn from lay users demonstrations. Since this data might be imperfect or contain errors, my focus is on developing methods for robots to learn safe and reliable behaviors from it.

Self Supervised Detection of Incorrect Human Demonstrations: A Path Toward Safe Imitation Learning by Robots in the Wild

We propose a Behavior Cloning for Error Detection (BED) framework that can detect incorrect human demonstrations in a self-supervised manner. With lay users demonstration we show that using our method robots can learn to avoid unsafe behaviors. We demonstrate the effectiveness of our method in RoboSuite simulation and with a Sawyer robot.

Self-Supervised Visual Motor Skills via Neural Radiance Fields

We propose a novel network architecture for visual imitation learning that exploits neural radiance fields (NeRFs) and key-point correspondence for self-supervised visual motor policy learning.

Projects

These include coursework, side projects and unpublished research work.

Kiddo: An educational robot for children

Simulation and hardware.

Lee: A biped walking robot

We developed a low cost full body biped walking robot. YouTube

Ribo Robot

A 24 DOF humanoid robot with a friendly interface capable of hand and arm manipulation.

Design and source code from Jon Barron's website